Intelligent metalworking control method

Abstract

The present invention relates to a metalworking automation system, and more specifically, to an intelligent metalworking control device and method that improves productivity and quality by comprehensively analyzing heterogeneous sensor data collected in real time during a cutting process performed on a machine tool, predicting abnormal conditions of the process in advance, and dynamically controlling processing conditions.

Description

Technology Field

[0001] The present invention relates to a metalworking automation system, and more specifically, to an intelligent metalworking control device and method that improves productivity and quality by comprehensively analyzing heterogeneous sensor data collected in real time during a cutting process performed on a machine tool, predicting abnormal conditions of the process in advance, and dynamically controlling processing conditions. Background Technology

[0002] Recently, the use of automated machine tools, such as Computer Numerical Control (CNC) machining centers, has become commonplace in the manufacturing industry to improve production efficiency and precision. These machine tools cut metal by moving the tool along a predetermined path and speed according to pre-programmed NC code.

[0003] However, most conventional metalworking systems operate using an open-loop control method, which limits their ability to actively respond to various abnormal situations that occur during the process. For example, unexpected micro-breakage of tools, chatter vibration caused by fluctuations in cutting load, and degradation of cutting fluid performance lead to serious quality defects, such as dimensional errors or surface defects in the workpiece.

[0004] To address these issues, methods such as periodically replacing tools based on processing time or usage frequency, or triggering alarms when the Root Mean Square (RMS) value exceeds a specific threshold, are being used in some systems. However, these methods fail to accurately reflect the actual condition of the tools, leading to the waste of tools with remaining lifespan or requiring reactive responses only after problems have become severe; consequently, they have been insufficient to fundamentally prevent sudden defects. Furthermore, despite the fact that processing instability is caused by complex factors—including not only mechanical vibration but also electrical instability and operator skill levels—conventional technologies have the drawback of failing to comprehensively consider these diverse factors. The problem to be solved

[0005] The present invention was devised to solve the problems of the prior art as described above, and aims to solve the following problems.

[0006] The first objective is to move away from fragmentary condition monitoring methods that rely solely on tool usage time or single physical quantities (such as vibration magnitude) and to provide a new technical means for comprehensively determining process conditions by organically combining multifaceted and heterogeneous data, including tool micro-geometry, history, data reliability, machining stability, coolant status, and operator factors.

[0007] The second objective is to provide a preventive control method that fundamentally prevents sudden production stoppages and the occurrence of defective products by predicting signs of quality defects or system failures based on collected data before they occur, rather than a reactive approach of responding after a problem arises, and by preemptively controlling processing conditions in real time.

[0008] The third task is to provide a dynamic control system that, rather than a static system relying solely on fixed thresholds or rules, can actively adapt to various processing environments and changes in conditions and improve judgment accuracy over the long term by continuously learning actual production results and optimizing judgment criteria autonomously.

[0009] The problems of the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0011] According to an embodiment of the present invention for solving the above problem

[0012] A data collection unit that collects metal processing process data including multiple sensors;

[0013] A data processing module that processes the above-mentioned collected process data;

[0014] A state prediction module that determines a plurality of process states, including abnormal tool conditions, machining stability risks, cutting fluid contamination conditions, and operator-cognitive load conditions, based on the above-mentioned processed process data, and predicts the final machining quality by synthesizing the determined plurality of process states;

[0015] A machining parameter generation module that generates control commands to modify real-time machining conditions according to the above-determined process state and predicted machining quality; and

[0016] A device control unit that transmits the above-mentioned generated control command to a machine tool;

[0017] Includes,

[0018] The above state prediction module is,

[0019] A step of calculating tool micro-profile drift by comparing the shape data of the tool edge acquired from the 3D laser scanner of the data acquisition unit with the initial shape data stored in the database;

[0020] A step of calculating a tool batch-lot freshness index by querying manufacturing lot information and history information of the tool from a database; and

[0021] A step of calculating digital log synchronization consistency by measuring timestamp errors between multiple sensors within the data collection unit;

[0022] Determine the abnormal condition of the tool, including,

[0023] When the digital log synchronization consistency exceeds a preset first threshold, it is determined to be in a 'warning' state requiring tool replacement as a top priority, regardless of the state of the tool micro-profile drift and the tool batch-lot freshness index, thereby preventing misjudgment due to a decrease in the reliability of the measurement data itself.

[0024] The above state prediction module is,

[0025] A step of calculating machining chatter phase synchronization by performing a cross-correlation analysis between vibration data received from first and second acceleration sensors, respectively attached to the spindle housing and the workpiece fixing jig and included in the data collection unit;

[0026] A step of calculating power distortion-phase skew by receiving voltage and current waveforms supplied to the spindle motor from a power quality analyzer included in the data acquisition unit and analyzing the phase difference thereof; and

[0027] A step of calculating a process noise spectrum anomaly index by calculating the difference between the frequency spectrum of the process noise collected from the microphone of the data collection unit and the previously stored steady-state spectrum;

[0028] Characterized by determining processing stability risk including,

[0029] The above state prediction module is,

[0030] If the power distortion-phase skew calculated above exceeds a second threshold and simultaneously the process noise spectrum anomaly index calculated above exceeds a third threshold, even if the processing chatter phase synchronization is below a 'stable' state threshold, it is determined as an 'initial unstable' state by judging it as an early sign that it may develop into a serious problem;

[0031] The above processing parameter generation module generates a spindle speed modulation command that continuously changes the spindle rotation speed within a certain range according to the result of determining the above 'unstable initial' state, thereby

[0032] It is characterized by preemptively avoiding the conditions for generating self-excited vibrations before mechanical chatter occurs through a combination of electrical instability and acoustic abnormal signals,

[0033] The above data processing module is,

[0034] For all data received from the above data collection unit, the line number (N-code) of the currently running NC program received from the computer numerical control unit (CNC) is matched, a high-precision timestamp based on International Standard Time (UTC) is assigned, and the data is stored in a time-series database;

[0035] The above device control unit is,

[0036] By converting the control commands generated from the above-mentioned machining parameter generation module into data packets supporting MTConnect or OPC-UA (Open Platform Communications Unified Architecture) standard communication protocols and transmitting them to the machine tool,

[0037] It is characterized by ensuring compatibility with heterogeneous machine tools and ensuring that all control operations are accurately linked to specific processing steps and time information.

[0038] The above state prediction module is,

[0039] By receiving feedback on actual past processing results and using a machine learning-based reinforcement learning algorithm, automatically adjust threshold values ​​to determine a 'caution' or 'warning' state,

[0040] If a defect occurs during the final quality inspection despite processing in a 'caution' state, the judgment accuracy can be improved through actual production results by lowering the threshold of the judgment criterion that caused the 'caution' state in a conservative direction during the next processing. Effects of the invention

[0042] According to the present invention for solving the above-mentioned problems, the following effects can be expected.

[0043] This invention provides a groundbreaking improvement in productivity and quality stability. It detects various abnormal signs that are the root causes of quality defects—such as micro-tool breakage, initial chatter, and cutting fluid performance degradation—at an early stage, and automatically performs preemptive measures, such as feed rate adjustment, spindle speed modulation, and tool replacement, before defects occur. Consequently, it fundamentally prevents the occurrence of unexpected defects that were unpredictable in conventional technology and maximizes the stability of unmanned automated processes, enabling the continuous production of uniform, high-quality products.

[0044] It offers the benefit of reducing operating costs. By precisely tracking the actual condition of tools and optimizing replacement timing, the waste of discarding usable tools can be reduced. Furthermore, losses such as material costs, processing time, and rework costs resulting from the production of defective goods can be minimized. Moreover, by preventing excessive loads on key system components (spindles, feed systems, etc.) in advance, it extends the lifespan of the equipment and reduces maintenance costs.

[0045] This invention enhances the intelligence and autonomy of the system. Rather than relying solely on fixed rules, the invention receives feedback on the quality data of actual produced products to autonomously learn and optimize judgment criteria (thresholds). This implies that the system becomes capable of making more sophisticated and accurate judgments over time through operational experience. Consequently, it is possible to implement a highly autonomous production system that can actively adapt to the processing of various materials and complex shapes while minimizing operator intervention.

[0046] It provides data-driven process analysis and optimization effects. This invention accumulates a vast amount of structured time-series data collected throughout the entire processing process. This data is not merely used for real-time control but can also be utilized as a valuable asset for analyzing and improving the entire process, such as by analyzing the root causes of specific defects or deriving optimal processing conditions for new materials.

[0047] The effects according to the present invention are not limited to those exemplified above, and a wider variety of effects are included within the present invention. Brief explanation of the drawing

[0049] Figure 1 illustrates an overall relationship diagram according to the present invention. Figure 2 illustrates a flowchart between all components according to the present invention. FIG. 3 illustrates a flowchart for determining a tool abnormal state according to the present invention. Figure 4 illustrates a flowchart for determining processing stability risk according to the present invention. Figure 5 illustrates a flowchart for determining the accumulation of cutting fluid contamination according to the present invention. Figure 6 illustrates a flowchart for predicting processing quality judgment according to the present invention. Specific details for implementing the invention

[0050] Hereinafter, various embodiments are described in more detail with reference to the attached drawings. The embodiments described in this specification may be modified in various ways. Specific embodiments may be depicted in the drawings and described in detail in the detailed description. However, specific embodiments disclosed in the attached drawings are intended only to facilitate understanding of various embodiments. Accordingly, the technical concept is not limited by specific embodiments disclosed in the attached drawings, and it should be understood that it includes all equivalents or substitutions that fall within the spirit and scope of the invention.

[0051] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but these components are not limited by the aforementioned terms. The aforementioned terms are used solely for the purpose of distinguishing one component from another.

[0052] Functions related to artificial intelligence according to the present disclosure are operated through a processor and memory. The processor may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. The one or more processors control the processing of input data according to predefined operation rules or artificial intelligence models stored in memory. Alternatively, if the one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0053] The predefined rules of operation or artificial intelligence models are characterized by being created through learning. Here, being created through learning means that a basic artificial intelligence model is trained using a number of training data by a learning algorithm, thereby creating predefined rules of operation or artificial intelligence models configured to perform desired characteristics (or objectives). Such learning may be performed on the device itself where the artificial intelligence according to the present disclosure is executed, or it may be performed through a separate server and / or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above.

[0054] An artificial intelligence model can be composed of multiple neural network layers. Each of the multiple neural network layers has multiple nodes and weight values, and performs neural network operations through calculations between the results of previous layers and the multiple weights. The multiple weights possessed by the multiple neural network layers can be optimized based on the learning results of the artificial intelligence model. For example, multiple weights can be updated so that the loss value or cost value obtained by the artificial intelligence model during the learning process is reduced or minimized. Additionally, to minimize the loss value or cost value, multiple weights can be updated in a direction that minimizes the gradient associated with the loss value or cost value. Artificial neural networks may include deep neural networks (DNNs), such as Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), Bidirectional Recurrent Deep Neural Networks (BRDNNs), or Deep Q-Networks, but are not limited to the examples mentioned above.

[0055] A network is a network that serves as a transmission path for web pages; it may be a closed network such as a LAN (Local Area Network) or WAN (Wide Area Network), but it is desirable for it to be an open network such as the Internet. The Internet refers to a global open computer network structure that provides the TCP / IP protocol and various services existing at its upper layers, namely HTTP (HyperText Transfer Protocol), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System), SMTP (Simple Mail Transfer Protocol), SNMP (Simple Network Management Protocol), NFS (Network File Service), and NIS (Network Information Service).

[0056] Terminals can be implemented in various forms. For example, the terminals described in this specification may include mobile terminals such as smartphones, tablet PCs, PDAs, portable multimedia players, and MP3 players, as well as fixed terminals such as smart TVs and desktop computers.

[0057] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. When a component is described as being “connected” or “connected” to another component, it should be understood that it may be directly connected to or connected to that other component, or that there may be other components in between. On the other hand, when a component is described as being “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

[0058] Meanwhile, a "module" or "part" for a component as used in this specification performs at least one function or operation. Furthermore, a "module" or "part" may perform a function or operation by hardware, software, or a combination of hardware and software. Additionally, a plurality of "modules" or a plurality of "parts," excluding a "module" or "part" that must be performed on specific hardware or on at least one processor, may be integrated into at least one module. A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0059] In addition, power, power transmission, and control therefor for the following assembly configurations and embodiments, including "by control," follow conventional technology including terminals, applications, hardware control modules, etc., so they are omitted to avoid redundancy.

[0060] In addition, the operation embodiments and configurations described in a general manner without being explained in detail below follow the prior art and are omitted in order to focus on describing the purpose of the present invention and the resulting effects.

[0061] Furthermore, in describing the present invention, if it is determined that a detailed description of related known functions or configurations may unnecessarily obscure the essence of the invention, such detailed description is abbreviated or omitted.

[0062] The metal processing control method according to the present invention can be implemented through a metal processing control device comprising a data collection unit (100), a data processing module (110), a state prediction module (120), a processing parameter generation module (130), and a device control unit (200).

[0063] The data acquisition unit (100) is defined as a set of hardware including a plurality of sensors and data acquisition devices that measure and collect physical data related to the process state in real time from major parts of the machine tool, such as the tool, workpiece, spindle, and cooling water supply system, while the metal processing process is being performed. The vibration sensors of the data acquisition unit (100) are attached to the spindle housing and the workpiece fixing jig, respectively, to measure the acceleration values ​​of mechanical vibrations occurring during processing in three axes, and this provides raw data for calculating the 'process chatter phase synchronization' of Process 2. The power quality analyzer of the data acquisition unit (100) is connected to the power supply line of the spindle motor to measure voltage, current waveforms, and phase difference, and this provides raw data for calculating the 'power distortion-phase skew' of Process 2. The 3D laser scanner of the data collection unit (100) is installed near the tool changer or within the processing area to scan the 3D shape data of the tool edge immediately after tool replacement or before the start of a specific process, which provides raw data for calculating the 'tool micro-profile drift' of Process 1. The ultrasonic sensor of the data collection unit (100) is attached to the outer wall of the main cutting fluid supply pipe to measure the bubble concentration inside the fluid, which provides raw data for calculating the 'cutting fluid microbubble index' of Process 3. In addition, a temperature sensor, an acoustic emission sensor, a microphone, a fine dust sensor, a closed-circuit television camera, etc., are installed at specific locations according to their respective purposes to measure the corresponding physical quantities. Specifically, the microphone may be installed near the cutting fluid injection nozzle or at the bottom of the spindle head to minimize the influence of external noise and clearly collect acoustic emissions from the cutting point. Furthermore, the fine dust sensor may be installed above the open part of the cutting fluid tank to effectively measure contaminant particles falling from the atmosphere.The above-mentioned closed-circuit television camera is installed with a field of view that includes the operator's face and hands, as well as the control panel of the machine tool, so that it can simultaneously detect the operator's gaze deviation and manual operation. Each sensor included in the data collection unit (100) converts the measured analog signal into a digital signal through a data acquisition device and automatically transmits it to a data processing module (110) via a wired Ethernet or wireless network in a real-time streaming manner, including a high-precision timestamp. At this time, the 'high-precision timestamp' has a precision in the millisecond (ms) range, and all data acquisition devices included in the data collection unit have a built-in 'time synchronization client' that periodically synchronizes time with a central NTP (Network Time Protocol) server. Through this, all data collected from different sensors share a unified time standard, which serves as a technical prerequisite for accurately calculating the 'digital log synchronization consistency' described later and reliably analyzing the causal relationship between data.

[0064] The data processing module (110) is defined as a software module that processes large volumes of raw data received from the data collection unit (100) into a form suitable for subsequent analysis and stores and manages them in a structured database. To process the raw data into a 'form suitable for subsequent analysis,' the data processing module performs a feature extraction algorithm that is appropriate for the characteristics of each data. For example, for time series data received from a vibration sensor, it calculates multiple statistical and frequency-based features, including not only simple RMS values ​​but also kurtosis (indicating the kurtosis of the data), crest factor (indicating the prominence of peak values), and energy values ​​of major frequency components obtained through frequency analysis. These multidimensional feature data provide a foundation that enables the subsequent state prediction module to make more sophisticated and accurate judgments. The data processing module (110) performs a digital filtering function to remove electrical noise or abnormal signals caused by the external environment included in the received data; for example, in the case of vibration data, it applies a band-pass filter that allows only signals of a specific frequency band to pass through, considering the frequency characteristics of the machine. In addition, the data processing module (110) synchronizes the filtered data with specific events of the processing process, and, for example, receives the line number and tool number of the currently running program from a computer numerical control unit to match which processing stage each data occurred in. Finally, all collected data is linked with a high-precision timestamp based on international standard time, a sensor identifier, a measurement value, and the processing program line number at that time, and is stored in a time-series database in a structured form according to a data schema such as the [Table] below to construct big data.

[0065] [graph]

[0066]

[0067] The data processing module (110) receives raw data from the data collection unit (100), provides processed and structured data in the form of a database query response according to the request of the state prediction module (120), and, in particular, in cases such as processing stability risk assessment (process 2) where real-time response is required, directly transmits real-time data through an internal memory bus to minimize latency.

[0068] The state prediction module (120) is defined as a core software module of the present invention that receives processed data from the data processing module (110), performs a plurality of predefined logical processing procedures to determine the current state of the process and predict future risks. The state prediction module (120) includes a plurality of functionally independent sub-modules, the first sub-module determines the abnormal state of the tool as 'normal', 'caution', or 'warning' grades as described in Process 1, the second sub-module determines the processing stability risk as 'stable', 'initial instability', or 'risk' grades as described in Process 2, the third sub-module determines the cutting fluid contamination state as 'good', 'performance degradation', or 'replacement required' grades as described in Process 3, and the fourth sub-module predicts the final processing quality as 'best', 'good', 'degradation expected', or 'defect risk' grades by synthesizing the determination results of the first to third sub-modules as described in Process 4. The state prediction module (120) receives data necessary for analysis from the data processing module (110) and transmits the judgment and prediction results derived from each sub-module to the machining parameter generation module (130). Additionally, the sub-modules interact to improve the dynamic adaptability of the system. For example, if the tool abnormality status is determined to be 'caution' grade in the first sub-module, this information is transmitted to the second sub-module to temporarily adjust the threshold value for determining machining stability risk to a more sensitive direction. This reflects the engineering fact that a worn tool is more vulnerable to chatter vibration in the system logic, and is characterized by increasing the reliability of the overall judgment by organically linking each state rather than judging it independently.

[0069] The machining parameter generation module (130) is defined as a software module that generates specific control commands to modify current machining conditions in real time or take preventive measures based on the judgment and prediction results received from the state prediction module (120). When the machining parameter generation module (130) receives a 'warning (replacement needed)' state from the state prediction module (120), it generates a tool change M-code command string that can be recognized by the computer numeric control unit. Additionally, when it receives an 'initial instability' state, it generates a data packet containing the spindle speed override register address of the computer numeric control unit and a target modulation range to continuously change the spindle rotation speed within a certain range. When it receives a 'degradation expected' state, it generates an intermediate inspection instruction text message to be displayed on the human-machine interface screen for the operator. If multiple state judgment results are received simultaneously, the machining parameter generation module determines the final command according to a rule that assigns the highest priority to commands related to safety. For example, if an 'instability initial' state (action: speed modulation) and a 'warning (replacement needed)' state (action: process stoppage and replacement) occur simultaneously, only the 'process stoppage and replacement' command, which is the action for the 'warning' state with a higher risk, is generated to ensure the safety of the system as the top priority. The processing parameter generation module (130) receives a state determination result from the state prediction module (120) and transmits the generated specific control command data to the device control unit (200).

[0070] The above device control unit (200) is defined as an interface hardware unit that receives a logical control command generated by the machining parameter generation module (130), converts it into an electrical signal or communication protocol that can be physically recognized and executed by the computer numerical control unit or programmable logic controller of the actual machine tool, and transmits it. The above device control unit (200) includes an internal 'command conversion map'. This map defines rules for converting a logical command, such as 'spindle speed modulation' received from the machining parameter generation module, into a specific execution command according to the type of connected machine tool. For example, for the same 'spindle speed modulation' command, the method is to convert it to "periodically write the target speed value to a specific node using the OPC-UA protocol" for Company A's CNC, and to "overwrite the value of system macro variable #510 via RS-232C communication" for Company B's older CNC. Through this, it is possible to respond to the diversity of hardware interfaces without changing the upper logic. The above device control unit (200) may be implemented as an industrial personal computer or gateway device that supports a standardized machine tool communication protocol, and transmits control commands by communicating directly with the latest computer numerical control system via an Ethernet network through the standard protocol. To ensure compatibility with older equipment, control commands are transmitted indirectly by sending physical signals by controlling specific input / output contacts of a programmable logic controller or by externally modifying user macro variable values ​​of the computer numerical control device via serial communication. The above device control unit (200) receives control command data from the machining parameter generation module (130), converts it into a physical signal or communication protocol data packet suitable for the machine tool, and automatically transmits it to the computer numerical control device and the programmable logic controller.

[0071] To explain the organic operation of the metal processing control device according to the present invention through a scenario, while the processing process is in progress, the vibration sensor and power analyzer of the data collection unit (100) transmit the detected raw data to the data processing module (110). The data processing module (110) processes the received data and provides it to the state prediction module (120). The second sub-module of the state prediction module (120) analyzes the data, determines the state as 'initial instability,' and transmits the result to the processing parameter generation module (130). The processing parameter generation module (130) generates a spindle speed modulation command data packet based on the 'initial instability' state and transmits it to the device control unit (200). The device control unit (200) converts the received data packet into a communication protocol format of the machine tool and transmits it to the computer numerical control device. The computer numerical control device receives the command and finely changes the spindle rotation speed in real time, thereby resolving processing instability and preventing quality degradation in advance.

[0072] Specifically, the main technical concept of the present invention is to go beyond merely monitoring the physical state of a machine in a metalworking process and to organically combine multidimensional information, including mechanical and electrical states, the state of consumables (cutting fluid), and human factors (operators), in order to predict the quality of the final product in advance and respond preemptively. The characteristic operation of the present invention to implement this includes the following step-by-step processing procedure.

[0073] First, the present invention collects heterogeneous data with different physical domains and time scales, such as the micro-geometry of a tool, machine vibration and power, cutting fluid bubbles and contamination, and operator behavior patterns, in real time and processes them into synchronized data for analysis. Next, based on the processed data, the abnormal condition of the tool (Process 1), the stability risk of the machining process (Process 2), and the contamination state of the cutting fluid (Process 3) are individually determined through independent specialized analysis logic. Subsequently, human factors, including the 'operator-cognitive load index' quantified by objective data, are organically combined with the determined multiple machine and material condition information. Through this, beyond the simple machine condition, the quality grade of the workpiece to be finally produced (from 'best' to 'risk of defect') is predicted in advance with high accuracy (Process 4). Finally, based on the predicted quality grade and individual status judgment results, optimal control commands (e.g., spindle speed modulation, intermediate inspection instructions, immediate process stoppage, etc.) are automatically generated to prevent quality degradation and directly intervene in the machine tool, thereby preemptively resolving the problem before defects actually occur.

[0074] The above-described Process 1 has as its core operation the determination of an abnormal condition of a tool by comprehensively evaluating three independent dimensions—the tool's physical shape, historical characteristics, and the reliability of measurement data—in order to overcome the limitations of conventional technology that judged the tool's condition based solely on fragmentary wear. To this end, the process first directly measures minute physical deformations at the tool's cutting edge using a laser scanner and simultaneously evaluates potential quality based on the tool's manufacturing lot and management history from a database. In particular, in addition to the physical and historical evaluations, the process analyzes the time synchronization error between each sensor data to include the reliability of the data itself as an independent judgment criterion. Finally, the results of the three different evaluations are combined according to a clear priority rule of "data reliability priority," thereby performing a multifaceted and highly reliable determination of the tool's abnormal condition, such as classifying a tool as dangerous if data reliability is low, even if its physical condition appears good.

[0075] Specifically, the state prediction module (120) of the present invention uses three different dimensions of judgment criteria—physical shape, historical data, and data reliability—to comprehensively determine the abnormal state of the tool, and each criterion acts complementarily to detect potential risks that cannot be identified by a single indicator. The state prediction module (120) uses tool micro-profile drift, which is a physical measurement value indicating the degree of deformation of the three-dimensional shape of the tool tip compared to the initial state or the previous measurement state, as a judgment criterion. This is intended to directly detect shape changes caused by micro-breakage or thermal deformation that are difficult to predict beyond simple wear levels, thereby most accurately capturing the actual cutting performance degradation of the tool. In addition, the state prediction module (120) uses the tool batch-lot freshness index, which is a normalized index calculated based on the tool's unique history information such as manufacturing lot information, storage period, number of re-sharpenings, and coating type, as a judgment criterion. This is based on the observation that even tools of the same specifications may have different inherent qualities depending on manufacturing deviations or management conditions, and aims to manage tools with a highly reliable history and those without by reflecting potential performance deviations of the tool that cannot be detected by physical sensors in the judgment. Furthermore, the state prediction module (120) uses the digital log synchronization consistency, which is the maximum time error value of data timestamps recorded between multiple sensors, programmable logic controllers, and computer numerical control devices within the data collection unit (100), as a judgment criterion. This is intended to increase the conservatism of tool state judgment when there is a high possibility that the data is contaminated, by setting the reliability of the data itself, which serves as the basis for judgment, as an independent judgment criterion, since even precise sensor data can cause serious errors in the analysis of causal relationships between data if the time axis is misaligned.

[0076] The logical processing procedure for making a final judgment by synthesizing the three aforementioned judgment criteria consists of an individual risk assessment stage that considers the characteristics of each criterion and a final judgment stage that weights and combines them. This is based on the technical grounds that abnormal conditions of a tool do not occur due to a single cause, and the influence of each cause on the result is not linear. For example, micro-breakage of a tool initially has a negligible impact on quality, but once it exceeds a certain threshold, it rapidly causes defects; furthermore, data synchronization errors pose a critical risk that can contaminate the judgment of the entire system, even at a minute level. To reflect the non-linear characteristics of each of these risks, the present invention adopts a method of combining a risk assessment logic differentiated for each criterion rather than simple summation.

[0077] The tool micro-profile drift is acquired by a non-contact 3D laser scanner installed inside the machine tool scanning the tool cutting edge before the tool change cycle or the start of each machining cycle. The scanned 3D point cloud data is compared with the initial 3D profile data stored in the system to calculate the distance value in micrometers of the point showing the largest deviation from the entire profile, for example, the maximum deviation relative to the initial profile is calculated as 5.2 micrometers. The tool batch-lot freshness index is obtained by acquiring a lot identifier from a barcode or radio frequency identification tag scanned by an operator when mounting a tool. The system uses this identifier as a key value to query an internal material management database to extract historical data such as the manufacturing date, number of regrindings, and coating material. It then calculates an index normalized to a value between 0 and 1 according to predefined rules, for example, the freshness index is calculated as 0.8 based on the conditions of one regrinding and a storage period of 12 months for a specific lot identifier. The above digital log synchronization consistency is such that all components of the system periodically synchronize time with a central network time protocol server, and the state prediction module (120) records the difference by comparing the timestamp included in the data with the reference time of the current central server whenever it receives data from each component, and calculates the largest value among all time error values ​​recorded during the last minute in seconds, for example, calculates the maximum time error between the computer numerical control unit and the vibration sensor during the last minute as 0.8 seconds.

[0078] The final result determined by the state prediction module (120) is classified into three states: 'Normal', 'Caution', and 'Warning', and each state is converted into a specific equipment control command by the machining parameter generation module (130). When determined to be in the 'Normal' state, a command is sent to proceed with the process while maintaining the currently set machining conditions without change. When determined to be in the 'Caution (condition adjustment)' state, a parameter override command is generated and automatically transmitted to the device control unit (200) to reduce the feed rate of the current machining path by a preset ratio and reduce the cutting depth in order to prevent deterioration of machining quality. When determined to be in the 'Warning (replacement needed)' state, a command is generated and automatically transmitted to stop the process and automatically replace the tool with a spare tool as soon as the currently ongoing machining cycle is safely completed, as it is determined that the tool can no longer be used.

[0079] For example, in a scenario immediately before starting the next machining cycle after high-speed machining high-hardness steel for 30 minutes using a tungsten end mill of a specific lot with a freshness index of 0.8, first, in the [Step 1] data acquisition step, a laser scanner scans the tool to measure the tool micro-profile drift value as 5.2 micrometers, queries the database for the tool batch-lot freshness index value as 0.8, and confirms the digital log synchronization consistency value as 0.8 seconds through timestamp monitoring. Next, in the [Step 2] individual risk assessment step, the measured tool micro-profile drift value of 5.2 micrometers exceeds the 'normal' threshold (3 micrometers) but is below the 'severe' threshold (10 micrometers), so the primary risk level is determined to be 'medium' grade; the retrieved tool batch-lot freshness index value of 0.8 is higher than the 'defective' threshold (0.5), so the secondary risk level is determined to be 'low' grade; and the verified digital log synchronization consistency value of 0.8 seconds exceeds the 'warning' threshold (0.5 seconds), so the tertiary risk level is determined to be 'high' grade. Next, in the [Step 3] comprehensive judgment step, the state prediction module (120) refers to a predefined judgment rule table and applies a rule to increase the comprehensive risk level even if other risk levels are low when the tertiary risk level is 'high', so the comprehensive risk level is determined to exceed the 'warning' threshold, and the final state is determined to be 'warning (replacement needed)'. Finally, in the [4] action execution step, the state prediction module (120) transmits a 'warning' state to the machining parameter generation module (130), and the machining parameter generation module (130) generates an automatic tool replacement command before the start of the next machining cycle and automatically transmits it to the device control unit (200).

[0080] If data reception from a specific sensor is interrupted for a certain period of time or longer, the sensor's measurement value is processed as 'unmeasurable,' and a command is transmitted to switch to the most conservative safety mode to prevent unexpected accidents. If contradictory data is detected, such as when physical deformation is judged to be 'low' while data reliability is judged to be 'very high,' the logic of the present invention prioritizes data reliability; therefore, even if the physical condition appears good, the data itself cannot be trusted, and a command is transmitted to forcibly correct the overall risk level to at least the 'Caution' grade.

[0081] When comparing the present invention with prior art management methods based on processing time or number of cycles, in normal wear scenarios, the prior art may waste tools with remaining life according to a set cycle; in contrast, the present invention tracks actual wear and extends the replacement cycle when life remains, thereby maximizing tool life and reducing costs. In scenarios involving unexpected micro-breakage, the prior art fails to detect this until the replacement cycle and takes reactive measures after processing defects occur; whereas, the present invention immediately detects abrupt changes in tool micro-profile drift values, determines a 'warning' state, and replaces the tool immediately, thereby preventing sudden defects and ensuring quality stability. In scenarios involving the use of defective lot tools, the prior art treats them the same as normal tools until physical wear appears; whereas, the present invention preemptively manages potential defect factors by determining a 'caution' state from the start due to a low tool batch-lot freshness index and managing processing conditions by lowering them. In scenarios involving sensor data synchronization errors, the prior art may make misjudgments based on incorrect sensor data; whereas, the present invention detects a decrease in data reliability through digital log synchronization consistency values ​​and blocks the root cause of judgment errors by withholding judgment or switching to a safe mode.

[0082] [Table 1]

[0083]

[0084] As a result of 10 simulation examples under various conditions, it was determined that the logic of the present invention operates consistently and reasonably by determining that when all input values ​​are within a good range, as in Examples 1 and 10, it is determined as 'normal'; when some values ​​are within a caution range, as in Examples 2 and 8, it is determined as 'caution (condition adjustment)'; and when physical deformation is severe, as in Examples 3 and 9, it is determined as 'warning (replacement needed)'. In particular, when the tool placement-lot freshness index is poor, as in Examples 4 and 6, it is determined as 'caution' or 'warning' in combination with other conditions, and when the digital log synchronization consistency value is poor, even if all other conditions are good, as in Examples 5 and 7, it is immediately determined as 'warning (replacement needed)'.

[0085] [Table 1-1]

[0086]

[0087] The judgment rule table used in the present invention sequentially examines conditions according to priority. The first priority rule immediately determines a 'Warning' and issues a replacement measure if digital log synchronization consistency exceeds a warning threshold, and the second priority rule determines a 'Warning' and issues a replacement measure if tool micro-profile drift exceeds a warning threshold. The third priority rule determines a 'Warning' and issues a replacement measure if tool micro-profile drift and tool batch-lot freshness index simultaneously exceed a caution threshold. The fourth and fifth priority rules determine a 'Caution' and issue a condition adjustment measure if each index individually exceeds a caution threshold, and the final sixth priority rule determines all other cases as 'Normal' and orders the maintenance of conditions.

[0088] The judgment rules presented in this invention are based on the technical necessity that stable production and the prevention of unexpected defects are possible only by simultaneously considering three fundamental risk factors affecting quality in the field of high-precision metalworking: actual physical damage, potential quality deviation, and data reliability issues. Since relying solely on a single criterion, as in the prior art, makes it impossible to respond to various types of unpredictable problems, the multi-criterion-based judgment rules of this invention constitute an objective and necessary configuration for solving technical challenges.

[0089] By adopting a physical criterion called tool micro-profile drift, the accuracy of tool life prediction is enhanced, and in particular, by detecting unexpected micro-breakage early, mass defects can be prevented in advance. Furthermore, by adopting a history criterion called the tool batch-lot freshness index, minute performance differences between tools of the same specification are reflected during the process, allowing for the preemptive management of the risk associated with potentially defective tools and enhancing machining stability. Finally, by adopting a reliability criterion called digital log synchronization consistency, the root causes of critical judgment errors based on incorrect data are blocked, thereby ensuring the reliability of the entire automation system.

[0090] The threshold values ​​used in this specification are not fixed constants, and their basis for existence and setting method are as follows. A threshold value is defined as a reference value for distinguishing system states such as 'Normal', 'Caution', and 'Warning'. The initial threshold value is automatically set by the system based on an accumulated experimental database, depending on the material to be processed and the type and diameter of the tool used. Additionally, during operation, the system periodically receives feedback on actual processing results, such as surface roughness and dimensional accuracy measurements of the processed part. If a defect occurs even when processing in a 'Caution' state, the system transmits a command to automatically adjust the relevant threshold value in a more conservative direction using a machine learning-based reinforcement learning algorithm.

[0091] The tool micro-profile drift (T_mt_caution), which is the 'caution' state threshold used in the present invention, is an engineering value inversely calculated from the quality requirements of the machined product. This is established through experimental data that proves the causal relationship with the actual machining results. The table below is an example of 10 simulation data sets measuring the maximum surface roughness (Rmax) of an actual workpiece according to changes in the tool micro-profile drift value for a specific tool (10-pipe 2-flute carbide end mill) and material (SKD11) combination.

[0092] [Table 1-2]

[0093]

[0094] According to the simulation data above, it can be observed that the probability of failure to meet quality requirements increases sharply starting from the point where the tool micro-profile drift value exceeds 3.0 μm. Therefore, the 'caution' threshold is set based on this experimental data, and the calculation logic is as follows.

[0095] Formula: T_mt_caution = Avg(Δ| P(Defect) ≥ 0.5) * α_safety

[0096] Here, Avg(Δ | P(defect) ≥ 0.5) represents the average micro-profile drift value at the point where the probability of defect occurrence is 50% or higher (approximately 3.15 μm in the table above).

[0097] α_safety means the safety factor (e.g., 0.95).

[0098] As a result, the threshold is set to approximately 3.0 μm, thereby securing objective grounds to define the state immediately before quality defects occur as a 'caution' state.

[0099] To verify how effective the control method of the present invention is compared to the prior art (time-based replacement) by applying a set threshold, 10 verification simulations were performed for various tool condition scenarios.

[0100] [Table 1-4]

[0101]

[0102] As a result of the verification simulation above, the prior art exhibited problems such as producing a total of 31 defective products in 4 out of 10 trials and wasting a total of 105 hours of tool life in 4 trials. In contrast, the present invention demonstrates a significant effect in preventing resource waste by fundamentally preventing the production of defective products through the detection of all premature failures in advance via a set threshold value and by maximizing the actual lifespan of the tool. This can be quantified by the following calculation formula.

[0103] Formula: Defect Prevention Rate = (1 - (Number of defects of the present invention / Number of defects of the prior art)) * 100 = (1 - (0 / 31)) * 100 = 100%

[0104] Formula: Resource Utilization Improvement Rate = (Additional Usage Time of the Present Invention / Wasted Time of Prior Art) * 100 = (103 hours / 105 hours) * 100 98.1%

[0106] Unlike conventional technology that judges the cause of processing instability based solely on the magnitude of vibration, the above-described Process 2 has as its core operation the organic combination of mechanical and electrical phenomena to trace the root cause of the problem and preemptively detect early signs before they develop into severe chatter vibration. To this end, the process first directly detects the intrinsic characteristics of chatter by analyzing the phase synchronization between multiple acceleration sensors, and simultaneously measures minute instability in the power supply that causes non-uniformity in spindle motor torque through a power quality analyzer. The most significant feature of this process is that it defines a complex pattern in which weak electrical instability signals and acoustic abnormal signals, which are individually not severe, occur simultaneously as an "initial instability" risk and detects it. When such early signs are detected, process stability is secured before severe chatter causing quality defects occurs by avoiding the conditions for self-excited vibration through a preemptive response that immediately modulates the spindle rotation speed slightly.

[0107] Specifically, the state prediction module (120) of the present invention uses judgment criteria derived from three different physical domains—mechanical vibration, electric power, and acoustic emission—to comprehensively evaluate the mechanical and electrical stability of the machining process and determine potential risks. Each criterion targets a different root cause that induces machining instability, and the combination thereof aims to detect signs of stability degradation at an early stage. The state prediction module (120) uses machining chatter phase synchronization, which is a correlation coefficient representing the real-time phase alignment between vibration signals measured by multiple acceleration sensors attached at different locations such as tools and workpieces, as a judgment criterion. This is intended to distinguish between general random vibrations occurring during machining and chatter, which is self-excited vibration that has a fatal impact on quality. By utilizing the phenomenon where vibrations from different parts synchronize to the same phase at a specific frequency when chatter occurs, the module precisely determines whether chatter has occurred by measuring the strength of this synchronization. In addition, the state prediction module (120) uses power distortion-phase skew, which is an electrical indicator indicating the degree to which the phase difference between the voltage waveform and the current waveform of the three-phase AC power supplied to the spindle motor deviates from the normal range, as a judgment criterion. This is intended to detect the impact of instability of the electrical supply system on mechanical stability as well as mechanical problems. Since phase skew of power can cause the torque of the spindle motor to become irregular, inducing minute vibrations and amplifying chatter, it tracks the fundamental electrical cause of processing instability.Furthermore, the state prediction module (120) uses a process noise spectrum anomaly index as a judgment criterion, which is a value that quantitatively indicates how different the frequency spectrum of real-time process noise collected by a broadband microphone installed inside the machine tool is from the noise spectrum pattern of a normal state stored in advance. This is intended to capture various types of process abnormal conditions that are difficult to detect by vibration or power alone, such as fine shaking in areas where vibration sensors are not attached, abnormal friction of cutting edges, or poor cooling water supply, through changes in sound. In particular, it increases the reliability of the stability judgment by detecting the characteristic that sharp noise occurs in a specific frequency band when chatter occurs.

[0108] The three criteria mentioned above have meaning independently, but a more precise judgment is possible through their combination. The processing procedure consists of an individual monitoring step that monitors each criterion in real time and a rule-based comprehensive judgment step that determines the final risk level by combining their states. This processing procedure is based on the technical basis that machining instability is caused by complex factors. For example, minor electrical instability alone is not a problem, but when coupled with the natural frequency of the tool, it can be amplified into severe mechanical chatter. Therefore, the present invention has a logical basis for capturing early signs of instability before they develop into serious problems by detecting not only cases where a single criterion exceeds a threshold, but also 'conditional combinations' where multiple criteria simultaneously exceed a 'caution' level threshold.

[0109] The above machining chatter phase synchronization simultaneously collects vibration data at a sampling rate of 1 kilohertz from a first acceleration sensor attached to the spindle housing and a second acceleration sensor attached to the workpiece fixing jig, and performs cross-correlation analysis in real time on the two collected time series data to calculate the maximum phase synchronization coefficient in a specific frequency band as a value between -1 and 1, for example, the maximum phase synchronization coefficient in the 350-hertz band is calculated as 0.85. The above power distortion-phase skew is calculated by a power quality analyzer connected to the spindle motor power supply line measuring the three-phase voltage and current waveforms in real time, calculating the voltage-current phase difference between each phase, and calculating the degree of imbalance between them as a single distortion value in degrees, for example, when the current phase difference is measured as 3.5 degrees compared to the steady-state phase difference, it is calculated that a skew of 2.0 degrees has occurred. The above process noise spectrum anomaly index is calculated by collecting noise with a microphone installed near the processing area, performing a Fast Fourier Transform on the collected acoustic signal to generate a frequency spectrum, comparing this with the average spectrum of the 'normal processing state' stored in the database, and calculating the difference between the two spectrum distributions as a single anomaly index in bits using the Kullback-Leibler divergence technique, for example, calculating the anomaly index of the current spectrum relative to the normal spectrum as 4.5 bits.

[0110] The determined final processing stability risk state is classified into three stages: 'stable', 'initial unstable', and 'dangerous', and the processing parameter generation module (130) generates real-time control commands accordingly. When determined to be in the 'stable' state, a command is transmitted to maintain the current processing conditions and operate all parameters normally. When determined to be in the 'initial unstable' state, it is determined that signs of instability are detected, even if not severe chatter, and a command is generated to finely modulate the spindle rotation speed within a range of plus or minus 5 percent of the current value to avoid conditions for self-excited vibration, and is automatically transmitted. When determined to be in the 'dangerous' state, it is determined that obvious chatter or system instability is detected, and to prevent quality defects, the feed rate is immediately reduced to 50 percent, the spindle rotation speed is lowered to a safe range, and a warning alarm is automatically transmitted to the operator.

[0111] [Table 2]

[0112]

[0113] For example, in a scenario where the cutting load increases as tool wear progresses during deep pocket machining of an aluminum block, first [Step 1] in the real-time data monitoring stage, the phase synchronization in the two acceleration sensors is maintained at a low level of 0.4, the phase skew in the power quality analyzer increases slightly to 1.8 degrees, and the noise anomaly index in the microphone increases slightly to 2.5 bits. Then, in [Step 2] in the individual state evaluation stage, the mechanical vibration state is determined to be 'stable' because the phase synchronization of 0.4 is below the 'danger' threshold (0.7), the electrical state is determined to be 'caution' because the phase skew of 1.8 degrees exceeds the 'caution' threshold (1.5 degrees) but is below the 'danger' threshold (3.0 degrees), and the acoustic state is determined to be 'caution' because the noise anomaly index of 2.5 bits exceeds the 'caution' threshold (2.0). Next, in the [3rd] comprehensive judgment step, the state prediction module (120) refers to the judgment rule table and determines that although none of the individual criteria are in a 'danger' state, the 'electrical state' and 'acoustic state' correspond to a combination rule where both satisfy the 'caution' grade, so it is judged as an early sign that could develop into a serious problem, and determines the final state as 'initial instability'. Finally, in the [4th] action execution step, the state prediction module (120) transmits the 'initial instability' state to the machining parameter generation module (130), and the machining parameter generation module (130) generates a spindle speed modulation command that continuously changes the current spindle rotation speed between 7600 and 8400 RPM and automatically transmits it to the device control unit (200).

[0114] If the microphone detects external impact noise rather than process noise and the noise anomaly index increases rapidly, the logic of the present invention includes a filtering logic that confirms there is no significant change in vibration and power data, determines this as a temporary external factor, and does not generate a warning; and if the signal becomes saturated because the measurement range of the acceleration sensor is exceeded due to severe chatter, the system considers this as the most severe level of 'danger' rather than 'unmeasurable' and transmits a command to immediately execute emergency measures.

[0115] When comparing the present invention with the prior art vibration magnitude-based management method, in the initial chatter occurrence scenario, the prior art fails to detect the vibration because the overall vibration magnitude is not yet large, and only recognizes it after chatter marks appear on the surface; in contrast, the present invention immediately detects chatter through a rapid increase in phase synchronization and determines it to a 'dangerous' state, thereby fundamentally preventing quality defects caused by chatter. In the micro-vibration scenario caused by electrical instability, the prior art judges it to be normal because the vibration magnitude is small, resulting in unexplained deterioration of machining precision; whereas, the present invention detects an increase in power distortion and phase skew to determine it to be in an 'initial unstable' state and prevents the problem from worsening by addressing the root cause of the instability through spindle speed modulation. In the cutting edge micro-breakage scenario, the prior art is difficult to detect because vibration changes are minimal; whereas, the present invention detects anomalies through changes in specific high-frequency bands of the process noise spectrum and determines it to be in an 'initial unstable' state, overcoming the limitations of a single sensor through multi-sensor fusion.

[0116] The judgment rule table used in the present invention sequentially examines conditions according to priority. The first and second priority rules each determine the condition as 'dangerous' and issue a measure to rapidly reduce the feed rate if phase synchronization or power distortion-phase skew exceeds an individual risk threshold. The third priority rule determines the condition as 'initial instability' and issues a spindle speed modulation measure if power skew and noise anomaly index simultaneously exceed a caution threshold. The fourth and fifth priority rules each determine the condition as 'initial instability' and issue a spindle speed modulation measure if phase synchronization or noise anomaly index individually exceeds a caution or danger threshold. Finally, the sixth priority rule determines all other cases as 'stable' and commands the maintenance of the condition.

[0117] The judgment rules of the present invention are based on objective principles of processing physics and electrical engineering. Since chatter possesses a distinct physical characteristic known as 'phase synchronization' and motor torque is directly influenced by 'power phase stability,' these fundamental causes cannot be distinguished solely by the overall 'magnitude' of the vibration. Therefore, the rule system of the present invention, which establishes individual judgment criteria tailored to each physical phenomenon and detects scenarios where these criteria interact to amplify the problem, is an essential and objective technical configuration for resolving complex processing instability issues.

[0118] By adopting the machining chatter phase synchronization standard, it clearly distinguishes between general vibration and chatter, reducing unnecessary alarms and ensuring reliability by accurately detecting only dangerous vibrations that affect actual quality. Furthermore, by adopting the power distortion-phase skew standard, it extends the diagnosis of machining instability causes from the mechanical to the electrical realm, enabling preemptive responses to root causes that were previously difficult to identify. Finally, by adopting rule-based combination judgment, it allows for the early detection of potential risk situations where multiple signs of instability appear simultaneously—even if a single phenomenon is not severe in itself—enabling preventive measures to be taken before serious failures or mass defects occur.

[0119] The threshold values ​​used in this specification are defined as reference values ​​for distinguishing the stability status of the system, such as 'stable', 'initial instability', and 'risk'. The initial threshold value is automatically set by the system based on a 'machine profile' that includes the target machining precision, the age and rigidity of the machine tool, and the type of tool primarily used; for example, when machining high-precision parts, all threshold values ​​are set more sensitively. Additionally, the system continuously learns data patterns from the time when the 'risk' state occurred in the past, and if it learns that chatter occurs repeatedly after a specific pattern, the system automatically generates that pattern as a new 'initial instability' judgment rule or sends a command to lower the existing threshold value to improve prediction accuracy on its own.

[0120] The processing chatter phase synchronization (T_sync_risk), which is the 'risk' state threshold used in the present invention, is derived through a direct correlation analysis with the actual chatter occurrence phenomenon. The table below is an example of 10 simulation data tests that examined whether chatter marks occurred on the processed surface according to changes in the processing chatter phase synchronization value under specific processing conditions.

[0121] [Table 2-1]

[0122]

[0123] The above data clearly demonstrates a critical characteristic where the probability of chatter marks increases sharply around a phase synchronization value of 0.7. Therefore, the threshold setting is based on this experimental fact, and the operational logic is as follows.

[0124] Formula: T_sync_risk = Min(Γ| P(Chat) ≥ 0.9)

[0125] Here, Min(Γ| P(chatter occurrence) ≥ 0.9) represents the minimum phase synchronization value at which chatter marks occur with a high probability of 90% or more.

[0126] As shown in the table above, chatter occurred in all four of the four samples with a phase synchronization of 0.7 or higher, indicating that the probability of failure is very high in this range. Therefore, the threshold is set to 0.7 to establish an objective criterion for reliably detecting chatter.

[0127] To verify how effective the control method of the present invention, which applies a set threshold, is compared to the prior art (based on vibration RMS magnitude), verification simulations were performed on 10 scenarios in which chatter is possible.

[0128] [Table 2-2]

[0129]

[0130] As a result of the verification above, the prior art failed to detect initial chatter with a low RMS magnitude, causing four defects, and misidentified external shock vibrations as chatter, resulting in two unnecessary process stoppages. In contrast, the present invention demonstrated a significant effect in accurately detecting and suppressing all initial chatter and preventing false detections caused by external shocks, based on the essential characteristic of chatter known as phase synchronization.

[0131] Formula: Chatter Detection Accuracy = (Number of Correctly Detected / Total Number of Actual Chatter Occurrences) * 100

[0132] Prior art: (1 / 5) * 100 = 20%

[0133] The present invention: (5 / 5) * 100 = 100%

[0134] Formula: False positive rate = (Number of false detections / Total number of non-chatter situations) * 100

[0135] Prior art: (2 / 5) * 100 = 40%

[0136] The present invention: (0 / 5) * 100 = 0%

[0138] Process 3 described above overcomes the limitations of conventional technology that manages the state of cutting fluid by relying on periodic sampling or a single indicator. Its core operation involves simultaneously tracking and managing two key performance degradation factors with different time scales: short-term, rapidly changing physical problems and long-term, gradual chemical problems. To this end, the process first monitors the concentration of microbubbles, which are directly related to the cooling performance of the cutting fluid, in real time using an ultrasonic sensor to respond immediately to sudden performance degradation. At the same time, by periodically measuring the level of air contamination around the machine tool and accumulating (integrating) the values ​​over time, it manages the total amount of contamination introduced from the outside that causes the spoilage of the cutting fluid over the long term. Finally, by combining the real-time internal state and the accumulated external contamination state to predict the optimal replacement time and perform necessary measures, the process realizes predictive maintenance that minimizes resource waste and prevents unexpected production stoppages.

[0139] Specifically, the state prediction module (120) of the present invention uses two judgment criteria to comprehensively determine the performance degradation state of the cutting fluid: real-time changes occurring within the cutting fluid and contamination factors introduced from the external environment and accumulated over time. This is intended to increase the accuracy of the judgment by reflecting the fact that the state of the cutting fluid is determined by both instantaneous factors and long-term cumulative factors. The state prediction module (120) uses a cutting fluid microbubble index, which is a normalized index that measures the concentration and size distribution of microbubbles contained within the fluid passing through the cutting fluid supply line in real time, as a judgment criterion. This is intended to directly detect the real-time degradation state of cooling and lubrication performance, which is one of the most important functions of the cutting fluid. If the concentration of microbubbles increases, the thermal conductivity of the cutting fluid decreases rapidly, failing to effectively lower the temperature of the tool and workpiece, which is a direct cause of shortened tool life and reduced machining precision. In addition, the above-mentioned state prediction module (120) periodically measures the concentration of fine particles and oil mist floating in the atmosphere around the machine tool and uses the environmental particle-contamination accumulation index, which is the total amount accumulated from the start of the process to the present, as a judgment criterion. This is intended to track and manage contamination factors that are introduced from the outside through the cutting fluid open tank, etc., and cause long-term deterioration of the chemical properties of the cutting fluid. Since these external contaminants cause gradual but fatal performance degradation, such as promoting the decay of the cutting fluid, forming sludge, and clogging the filter, the expected lifespan of the cutting fluid is predicted and the replacement time is determined by managing this accumulated over time.

[0140] The processing procedure is structured to determine the final state of the cutting fluid by individually evaluating internal states that change in real time and external factors that accumulate over the long term, and then combining the two. This procedure is based on the technical premise that cutting fluid performance degradation occurs at two different time scales: microbubbles are a short-term issue that fluctuates rapidly from seconds to minutes depending on the machining load, whereas external contaminants are a long-term issue that accumulates gradually over hours to days. Since prior art has limitations in that it focuses primarily on a single aspect, failing to address short-term problems or predict long-term life degradation, the present invention adopts a logical configuration that combines real-time measurements with the concept of time integration to resolve issues across both time scales; this constitutes a technical necessity for accurately modeling the physical phenomena of cutting fluid performance degradation.

[0141] The above cutting fluid microbubble index is obtained by attaching an ultrasonic transducer to the outer wall of the main cutting fluid supply pipe. Ultrasonic waves are emitted, and the attenuation rate and phase change of the signal returning after scattering by microbubbles inside the fluid are analyzed to calculate the volume occupancy rate of bubbles in the fluid in real time. This value is normalized to an index between 0 and 1. For example, when the measured volume occupancy rate of bubbles is 3 percent, the normalized microbubble index is calculated as 0.45. The above environmental particle-contamination cumulative index is calculated by using a light scattering type fine dust sensor and an oil mist sensor installed on the top of the cutting fluid tank to measure atmospheric pollution concentrations at 1-minute intervals. The inflow amount during that time is calculated by multiplying the measured concentration value by the measurement time interval. This value is continuously added to the total cumulative value in the database and recorded. For example, when the fine dust concentration at the current time is 150 micrograms per cubic meter, the inflow amount for one minute is calculated and added to the existing cumulative value, resulting in the total cumulative index being updated to 8,500.

[0142] The identified cutting fluid status is classified into three stages: 'Good', 'Degraded', and 'Replacement Required', and automated measures are executed via a programmable logic controller accordingly. If the status is determined to be 'Good', a command is sent to operate the cutting fluid supply system normally. If the status is determined to be 'Degraded', it is determined that there is a temporary decrease in cooling performance due to an increase in microbubbles; a command is automatically sent to increase the pressure of the cutting fluid supply pump by 10 percent to forcibly expel bubbles from the fluid and to operate the automatic defoaming agent injection device once. If the status is determined to be 'Replacement Required', it is determined that the cutting fluid has reached the end of its lifespan due to accumulated contamination; an alarm is sent to the operator after the current machining is completed stating that a complete replacement of the cutting fluid and tank cleaning are required, and a command is automatically sent to set the next machining start to a locked state.

[0143] For example, in a scenario where titanium material is machined under high load on a machine tool that has been operated for 100 hours after the cutting fluid has been replaced, first, in the [Step 1] data acquisition and accumulation stage, the ultrasonic sensor detects bubbles that have increased due to the high load and measures the real-time microbubble index as 0.6, and the air pollution sensor confirms that the environmental particle-pollution accumulation index has reached 15,000 by steadily accumulating the inflow of external pollutants over the past 100 hours. Then, in the [Step 2] individual state evaluation stage, the real-time internal state is determined to be 'performance degradation' because the microbubble index of 0.6 exceeds the 'performance degradation' threshold (0.5), and the pollution accumulation index of 15,000 is determined to be 'long-term accumulation' because it exceeds the 'caution' threshold (10,000) but is below the 'replacement' threshold (20,000). Next, in the [3rd] comprehensive judgment step, the state prediction module (120) refers to the judgment rule table and determines the final state as 'performance deterioration' by defining it as a state where immediate replacement is not necessary but active measures are required, because it corresponds to a combination rule where the real-time state is 'performance deterioration' and the accumulated state is 'caution'. Finally, in the [4th] action execution step, the state prediction module (120) transmits the judgment result to the programmable logic controller, and the programmable logic controller immediately sends a signal to increase the cutting fluid pump pressure by 10 percent and executes a command to open the defoaming agent injection valve for 5 seconds.

[0144] If the microbubble index increases abnormally while the cutting fluid flow sensor value simultaneously decreases, it is determined that the issue is due to a supply pump failure or cavitation rather than a problem with the cutting fluid itself. In this case, a command is sent to execute exception handling logic that immediately halts machine operation and triggers a pump inspection alarm, rather than taking any cutting fluid-related measures. If the air pollution sensor value explodes abnormally within a short period, this is treated as a transient event; a command is sent to prevent distortion in the assessment of long-term trends by excluding the data from that time from the cumulative value calculation or reflecting it with a reduced weight.

[0145] When comparing the present invention with prior art methods such as periodic replacement or pH concentration management, in a scenario where bubbles increase temporarily due to high-load machining, the prior art fails to detect this, resulting in unexplained thermal deformation and accelerated tool wear; in contrast, the present invention detects a real-time increase in the microbubble index, determines it as 'performance degradation,' and immediately responds to the real-time degradation of cooling performance through measures such as increasing pump pressure, thereby protecting quality stability and tool life. In a scenario of gradual external contamination accumulation, the prior art fails to detect this until the replacement cycle, leading to sudden cutting fluid spoilage or filter clogging; whereas, the present invention continuously monitors the trend of life degradation through the contamination accumulation index, predicts the 'replacement needed' point, enables preventive maintenance, and prevents sudden system shutdown. In a low-load machining scenario in a clean environment, the prior art discards usable cutting fluid according to a set replacement cycle; whereas, since the contamination accumulation index remains low, the present invention automatically extends the replacement cycle to maintain a 'good' state, thereby preventing resource waste and reducing operating costs.

[0146] [Table 3]

[0147]

[0148] The judgment rule table used in the present invention sequentially examines conditions according to priority. The first and second priority rules each determine that "replacement is required" and issue a replacement alarm and system lock measures when the contamination accumulation index or the microbubble index exceeds an individual replacement threshold. The third priority rule determines that "performance degradation" occurs when both the microbubble index and the contamination accumulation index simultaneously exceed a caution threshold and issues measures to increase pump pressure and administer an antifoaming agent. The fourth priority rule determines that "performance degradation" occurs when the microbubble index individually exceeds a caution threshold and issues the same measures. Finally, the fifth priority rule determines all other cases as "good" and orders normal operation.

[0149] The judgment rule of the present invention is based on the objective principle that the performance of a cutting fluid is determined by two independent factors: the physical state of the internal fluid and chemical contaminants introduced from the outside. Since prior art had limitations in focusing on only one of these aspects, the present invention presents a judgment rule that quantitatively measures both of these factors and considers their interaction; this is an essential technical configuration to reflect all objective physical and chemical phenomena that cause the degradation of cutting fluid performance.

[0150] By adopting a microbubble index, it is possible to directly respond to changes in cooling performance in real time, which has the effect of protecting tools and workpieces from thermal damage and maintaining machining precision even during high-load machining. By adopting an environmental particle-contamination accumulation index, it enables predictive maintenance by scientifically predicting the expected lifespan of the cutting fluid and determining the optimal replacement time, thereby reducing wasteful replacements and preventing production stoppages caused by sudden spoilage. Furthermore, by simultaneously managing short-term and long-term problems through real-time and cumulative combined judgments, it has the technical effect of maximizing the overall efficiency and stability of cutting fluid management.

[0151] The threshold values ​​used in this specification are defined as standard values ​​for distinguishing the condition of the cutting fluid, such as 'Good', 'Degraded Performance', and 'Replacement Required'. The system sets the initial threshold value based on the type of cutting fluid used, the management standards recommended by the manufacturer, and the capacity of the cutting fluid tank. The system records the cutting fluid replacement cycle and stores the cumulative index value at the time of actual replacement in the database. If a problem occurs at a specific cumulative index on average over multiple replacement cycles, the system automatically lowers the 'Replacement Required' threshold based on past operational data and sends a command to optimize the judgment criteria to suit the field situation.

[0152] The environmental particle-contamination accumulation index (T_acc_replace), which is the 'replacement needed' threshold used in the present invention, is set based on data directly related to the actual performance degradation of the cutting fluid. The table below is an example of 10 simulation data sets tracking changes in bacterial colony-forming units (CFU / mL) measured in cutting fluid samples as the contamination accumulation index increases.

[0153] [Table 3-1]

[0154]

[0155] The above data shows a clear inflection point where the number of bacteria increases exponentially as the contamination accumulation index exceeds 20,000, surpassing the spoilage threshold (10^6 CFU / mL). The threshold is set by predicting the point just before actual spoilage occurs, and the calculation logic is as follows.

[0156] Formula: T_acc_replace = Value(Accumulated exponent @ d²(Number of bacteria) / d(Exponent)² = Max) * α_safety

[0157] Here, Value(...) refers to the cumulative exponent value corresponding to the inflection point (maximum second derivative) where the growth rate of the bacteria count is maximum. (Approximately 21,000 in the table above)

[0158] α_safety means the safety factor (e.g., 0.95).

[0159] As a result, the threshold is set to approximately 20,000, providing an objective standard for taking preventive measures before reaching the irreversible state of cutting fluid spoilage.

[0160] To verify how effective the control method of the present invention, with a set threshold applied, is compared to the prior art (periodic replacement), 10 verification simulations were performed on different contaminated environments.

[0161] [Table 3-2]

[0162]

[0163] As a result of the verification above, the prior art failed to respond to environmental changes, wasting a total of 7.5 months' worth of cutting fluid in a clean environment and causing a total of 4 production stoppages (a total of 4 days) in a contaminated environment. In contrast, the present invention demonstrated a significant effect of preventing all production stoppages and minimizing resource waste by flexibly adjusting the replacement timing based on the actual contamination state.

[0164] Formula: Total Operating Cost = (Cutting Fluid Replacement Cost * Frequency) + (Opportunity Cost of Downtime * Days)

[0165] Prior art: (Cost * 10 times) + (Opportunity cost * 4 days)

[0166] The present invention: (Cost * 10 times) + (Opportunity cost * 0 days)

[0167] As a result, it proves the effect of eliminating 100% of the opportunity cost caused by production stoppage.

[0169] Process 4 described above departs from the conventional view that quality management is merely a matter of individual process variables. Its core operation is to comprehensively predict the quality grade of the final product in advance by quantitatively including not only machinery, materials, and the environment, but also "human error" that was previously overlooked in automated systems. To this end, this process first synthesizes the conditions of tools, machining stability, and cutting fluid determined in the preceding processes. Additionally, it adds the "Worker-Clarity Load Index," objectively calculated through the frequency of manual operator intervention, alarm response times, and CCTV video analysis, as a key judgment criterion. By modeling a non-linear relationship where the probability of quality defects increases sharply when the combined risk factor—encompassing both mechanical and human factors—exceeds a specific threshold, this process determines the final quality grade from "Excellent" to "Defect Risk." Based on these prediction results, preemptive measures—such as ordering intermediate inspections or immediately halting the process before defects actually occur—are taken, thereby overcoming the fundamental limitations of post-inspection methods and contributing to defect-free production.

[0170] Specifically, the state prediction module (120) of the present invention uses four comprehensive judgment criteria, including machine, material, environment, as well as human factors, to predetermine the expected final quality grade when a specific processing process is completed. This is intended to maximize the accuracy of the prediction by reflecting the fact that processing quality is the result of complex system interactions rather than a single factor. The state prediction module (120) uses a tool defect comprehensive index, which is a risk index that comprehensively evaluates the physical deformation, history, and data reliability of the tool calculated in Process 1, as a judgment criterion. This is intended to comprehensively reflect the current state of the cutting tool, which has the most direct impact on quality. A high index indicates a high probability of defects occurring on the processing surface or dimensional errors occurring. Additionally, the state prediction module (120) uses the machining stability risk index, which is a risk index that comprehensively evaluates mechanical vibration, electrical stability, and acoustic patterns calculated in process 2, as a judgment criterion. This is intended to reflect the impact on quality caused by a decrease in the dynamic stability of the process, such as chatter. A high index indicates that vibration marks may remain on the machining surface or lead to sudden breakage of the tool. Furthermore, the state prediction module (120) uses the cutting fluid contamination accumulation index, which is a risk index that comprehensively evaluates the real-time performance and cumulative contamination of the cutting fluid calculated in process 3, as a judgment criterion. This is intended to reflect the indirect but fatal impact on quality caused by a decrease in cooling and lubrication performance. A high index indicates a high probability of dimensional accuracy degradation due to thermal deformation or surface scratching due to poor chip evacuation.Finally, the state prediction module (120) uses a worker-cognitive load index, which is a normalized index representing the current cognitive load or fatigue of the worker calculated by synthesizing the frequency of manual operation by the worker, the average response time to system alarms, and attention distraction patterns through closed-circuit television video analysis, as a judgment criterion. This is intended to quantitatively evaluate the impact of human intervention, such as handling unexpected situations or quality inspection, on the final quality, even in automated processes. A high index indicates an increased risk of setting errors, neglect of problematic situations, or omission of inspections caused by worker error.

[0171] The processing procedure consists of a step of calculating the four individual risk indices mentioned above into a single 'comprehensive risk factor' and a step of applying this comprehensive risk factor to a non-linear quality degradation model to determine the final quality grade. This processing procedure is based on the technical premise that quality defects in the processing stage exhibit non-linear characteristics, where defects do not appear even if various risk factors accumulate up to a certain level, but the probability of occurrence increases sharply the moment a specific threshold is exceeded. To model this 'quality breakdown' phenomenon, the present invention adopts a logical structure in which the quality grade changes rapidly according to the critical interval, rather than a simple linear summation method; this is an inevitable technical configuration that aligns with the defect occurrence patterns observed in actual industrial settings.

[0172] The above-mentioned comprehensive tool defect index, machining stability risk index, and coolant contamination accumulation index are obtained by receiving the results of processes 1, 2, and 3 in real time, and each index is internally converted into grades such as 'low', 'medium', and 'high' for use. For example, the current tool defect index is determined and obtained as 'medium', the stability risk index as 'low', and the coolant contamination index as 'medium'. The above-mentioned operator-cognitive load index is calculated by analyzing control panel operation logs to calculate the number of manual parameter modifications by the operator per unit time, measuring the average time until the operator presses the confirmation button when a system alarm occurs, and detecting the frequency of the operator deviating from a safe posture or their gaze leaving the equipment using a deep learning-based posture estimation model on closed-circuit television footage showing the machining area. These three factors are weighted and summed to normalize the result into an index between 0 and 1. For example, the final cognitive load index is calculated as 0.72 based on the weighted summation results according to conditions such as frequent manual operation, delayed alarm response, and frequent gaze departure.

[0173] The predicted final quality grade is classified into four levels: 'Excellent', 'Good', 'Expected Deterioration', and 'Risk of Defect', which are utilized for proactive measures. If the grade is determined to be 'Excellent' or 'Good', a command is sent to proceed with the current process as planned without taking additional actions. If the grade is determined to be 'Expected Deterioration', quality degradation is a concern but not critical, so instructions are automatically sent to the operator to perform an intermediate inspection after the current processing is completed and before moving on to the next process. If the grade is determined to be 'Risk of Defect', defects are predicted to occur with a high probability, so to prevent further waste of resources, the current process is immediately and safely halted, and a detailed diagnostic report on all risk factors is generated and automatically sent to the manager.

[0174] For example, in a scenario where a low-skilled worker is assigned to the final finishing stage of a mold core with a complex shape, first [Step 1] in the data aggregation stage, tool wear accumulates, so the tool defect index is rated 'medium', the machining stability risk index is rated 'low' because the load from the finishing process is low, the cutting fluid contamination accumulation index is rated 'low' because the cutting fluid condition is good, and the worker-cognitive load index is rated 'high' because the worker is not familiar with the process conditions. Next, in [Step 2] in the comprehensive risk factor calculation stage, the state prediction module (120) converts four individual risk grades into one comprehensive risk factor by applying predefined weights, and as a result, the comprehensive risk factor records a high value close to the 'risk zone' threshold. Next, in [Step 3] in the final quality grade determination stage, the calculated comprehensive risk factor is compared with the threshold of the judgment rule table, and since the comprehensive risk factor exceeds the 'expected deterioration' threshold but does not exceed the 'defect risk' threshold, the final predicted quality grade is determined as 'expected deterioration'. Finally, in the [4] action execution step, the state prediction module (120) transmits the judgment result to the worker interface and the central control system, and sends a command to automatically display an intermediate inspection guideline message on the worker monitor.

[0175] If closed-circuit television (CCTV) video analysis determines that an operator has been away from their post for longer than a set amount of time, a command is sent to forcibly set the operator-cognitive load index to the highest risk level. When starting the processing of a new product with no prior processing history, the system automatically sets the quality judgment threshold to be generally more conservative, and then sends a command to execute learning logic that gradually relaxes the threshold to ensure productivity as data from successfully processed products accumulates.

[0176] When comparing the present invention with the prior art method of final inspection after machining is complete, in the scenario of precision degradation due to tool wear, the prior art discovers defective products after machining is complete, resulting in product disposal and rework costs. In contrast, the present invention predicts a 'predicted degradation' grade in advance using a tool defect index and orders an intermediate inspection, thereby securing an opportunity for correction before defects occur, thus preventing the production of defective products at the source and reducing rework costs. In the scenario of setting errors caused by operator error, the prior art fails to recognize operator mistakes and analyzes the cause only after mass defects have occurred; whereas the present invention detects a surge in the operator-cognitive load index, predicts a 'risk of defect,' and immediately halts the process, thereby minimizing damage and preventing mass defect accidents. In the scenario of complex problem occurrence, the prior art cannot identify the problem through single-cause analysis, whereas the present invention provides a detailed diagnostic report by comprehensively analyzing all factors, including tools, stability, and operators, thereby enabling the rapid identification of the root cause of the problem and facilitating the establishment of measures to prevent recurrence.

[0177] [Table 4]

[0178]

[0179] The judgment rule table used in the present invention sequentially examines conditions according to priority. The first priority rule determines the predicted quality grade as 'risk of defects' and issues measures for immediate process suspension and reporting to a manager if the overall risk factor exceeds the critical risk threshold. The second priority rule determines the grade as 'expected deterioration' and issues a directive for an intermediate inspection after completion if the overall risk factor exceeds the warning threshold. The third and fourth priority rules determine the grade as 'good' or 'best,' respectively, if the overall risk factor exceeds the caution threshold or in all other cases, and issue measures for normal continuation.

[0180] The judgment rules of the present invention are based on the objective fundamental principle that the quality of a final product is determined by the complex interaction of the four major elements of production: machinery, materials, methods, and people. Since prior art had limitations in managing only some of these elements in a fragmentary manner, the present invention quantifies and integrates all key judgment criteria corresponding to these four elements to predict quality. This systematic implementation of the objective fundamental principles of quality control is an essential technical configuration necessary for accurate quality prediction.

[0181] By synthesizing multiple indices, the judgment results of individual processes are integrated under the single goal of final quality, enabling high-level decision-making that comprehensively grasps the operational status of the entire system and predicts final results. Furthermore, by adopting the worker-cognitive load index, the 'human factor' that was overlooked by existing automation systems is included in the quality management loop, thereby dramatically increasing the accuracy of predictions and having the unique effect of proactively managing potential risks that may arise during the collaboration process between automation and humans. Through prior prediction and action, it shifts the quality management paradigm from a reactive approach that responds after defects occur to a preventive approach that predicts and responds before defects occur, thereby fundamentally improving the efficiency and reliability of the entire production process.

[0182] The quality judgment thresholds used in this specification are defined as standard values ​​of comprehensive risk factors for distinguishing predicted quality grades such as 'best', 'good', 'expected deterioration', and 'risk of defect'. Initial thresholds are automatically set by the system according to the tolerance grade and surface roughness requirements for the product, and, for example, in the case of a product requiring high precision, all thresholds are set very conservatively. In addition, the system continuously maps the final inspection results of the actual produced product with the comprehensive risk factor values ​​at that time and stores them in a database. When the accumulated data exceeds a certain amount, it recalculates the correlation between the actual probability of defect occurrence and the comprehensive risk factor through machine learning techniques such as logistic regression analysis, and based on this, transmits a command to automatically update the optimal threshold for distinguishing each quality grade.

[0183] The 'risk of defect' judgment threshold (T_risk_critical) used in the present invention is set by statistically analyzing large-scale historical production data. The table below is an example of data showing the overall risk factors at the time of production and the final quality inspection results of 10 products produced in the past.

[0184] [Table 4-1]

[0185]

[0186] The above data shows a tendency for the probability of defects to increase as the overall risk factor increases. The present invention derives objective threshold values ​​by applying statistical modeling to such a large-scale dataset (actually over several thousand), and the computational logic is as follows.

[0187] Formula: P(Defect|X) = 1 / (1 + e^-(β+ β)

[0188] Here, P(Defect|X) is a logistic regression model representing the probability of a defect occurring given the overall risk factor X.

[0189] β_0 and β are the coefficients of a model trained on a large dataset.

[0190] The 'risk of defect' threshold T_risk_critical is set to the minimum X value satisfying P(Defect|X) ≥ 0.9. In other words, the point where the probability of a defect occurring is statistically predicted to be 90% or higher is defined as the 'risk of defect'. For example, if the trained model predicted P=0.91 when X=0.9, T_risk_critical is set to 0.9.

[0191] To verify how effective the pre-prediction method of the present invention, with a set threshold applied, is compared to the prior art (post-inspection), a verification simulation was performed on 10 scenarios with a possibility of defect occurrence.

[0192] [Table 4-2]

[0193]

[0194] As a result of the verification above, in a total of six defect occurrence scenarios, the prior art produced defective products in all cases, resulting in massive resource waste. On the other hand, the present invention demonstrated a significant effect in fundamentally preventing resource waste by detecting all defect occurrences in advance and stopping the process through accurate predictions based on a statistical model.

[0195] Formula: Defect Production Prevention Rate = (Number of defective products produced by the prior art - Number of defective products produced by the present invention) / Number of defective products produced by the prior art * 100

[0196] = (6 - 0) / 6 * 100 = 100%

[0197] This demonstrates that the present invention solves the fundamental problems of post-processing methods and can drastically reduce unnecessary production costs through a new paradigm of preventive quality control.

[0199] The overall operation process and organic interaction of the intelligent metalworking control method according to the present invention are described in detail step-by-step through a specific scenario involving complex situations as follows. The machining process assumes a situation requiring very high precision and surface roughness, specifically the final finishing stage of high-strength titanium alloy parts for aircraft. The machine tool is a 5-axis machining center, and in the initial state, the tool is equipped with a ball end mill that has been used approximately 80 percent compared to a new one, and the manufacturing lot of this tool has shown average performance based on past data. The operator is set to be immediately after a highly skilled day shift worker has left and been replaced by a relatively less experienced night shift worker, and the cutting fluid is set to have accumulated some degree of contamination after being used for a considerable amount of time since replacement.

[0200] When the finishing machining program starts, each sensor included in the data collection unit (100) simultaneously begins collecting real-time data, a 3D laser scanner scans the current shape of the tool, acceleration sensors of the spindle housing and jig collect vibration data, a power quality analyzer collects power data of the spindle motor, ultrasonic sensors and contamination sensors collect state data of the cutting fluid, and control panel operation logs and closed-circuit television collect operator behavior data. All of this raw data is transmitted in real-time to the data processing module (110) along with high-precision timestamps.

[0201] The data processing module (110) processes the received raw data and converts it into meaningful judgment criteria data that can be used by the state prediction module (120), calculates the tool micro-profile drift value as 2.8 micrometers by comparing the laser scan data with the initial profile, calculates the machining chatter phase synchronization value as 0.45 as a result of cross-correlation analysis of the vibration data, calculates the power distortion-phase skew value as 1.9 degrees by analyzing the power data, calculates the cutting fluid microbubble index value as 0.4 and the environmental particle-contamination accumulation index value as 16,000 by analyzing the cutting fluid data, and calculates the operator-cognitive load index value as 0.65 because it is detected that the operator feels anxious about the process speed, frequently manually operates the feed rate override, and responds late to intermittent system alarms.

[0202] The state prediction module (120) receives the above-determined judgment criteria data and performs state determination and prediction in parallel within each internal sub-module. First, according to process 1, the tool abnormality state is determined. Since the drift value of 2.8 micrometers is close to the good limit, the freshness is average, and the synchronization consistency is good, it is determined to be in a comprehensive 'normal' state according to the rule table. Next, according to process 2, the machining stability risk is determined. Since the phase synchronization is stable but the phase skew exceeds the 'caution' threshold, it is determined to be in an 'unstable initial' state according to the rule table. Subsequently, according to process 3, the cutting fluid contamination state is determined. Since the real-time state is good but the long-term accumulated contamination is significant, it is determined to be in a 'good' state according to the rule table. Finally, the predicted machining quality is determined according to process 4. Since the tool condition is normal but the stability condition is in the early stages of instability, the coolant condition is good, and the operator-perceptual load is high, the weighted sum of the four conditions has a significant impact, particularly due to the high risk of the operator-perceptual load index. Consequently, the overall risk factor exceeds the 'expected deterioration' threshold, and the final predicted quality grade is determined as 'expected deterioration'.

[0203] The processing parameter generation module (130) combines multiple judgment results received from the state prediction module (120) to simultaneously generate multiple control commands most suitable for the current situation. In response to the 'initial instability' judgment, it generates a command data packet to modulate the spindle speed within a plus or minus 5 percent range to suppress chatter, and in response to the 'expected deterioration' judgment, it generates an intermediate inspection guideline text message to measure the dimensions of three key parts after processing is completed because quality deterioration is predicted, and does not generate separate control commands for other items judged as 'normal' and 'good'.

[0204] The device control unit (200) receives two generated control commands and transmits each to the machine tool in an appropriate manner. It converts the received spindle speed modulation command data packet to match the machine tool's communication protocol and transmits it to a computer numerical control unit, and the computer numerical control unit immediately executes this command to begin changing the spindle rotation speed in real time, thereby preventing electrical instability from being amplified into mechanical chatter. In addition, it transmits the received intermediate inspection instruction text message to a human-machine interface screen for the operator, and after the machining is completed, the operator measures the part according to the instructions displayed on the screen and inputs the results into the system, thereby providing an opportunity to detect and correct minute dimensional defects that could have occurred due to the operator's lack of skill at an early stage. Specifically, the device control unit (200) receives two generated control commands. Here, since the 'spindle speed modulation' command for the 'initial instability' state and the 'intermediate inspection instruction' command for the 'expected degradation' state do not conflict with each other, the device control unit converts both commands in a manner suitable for the corresponding machine tool and executes them simultaneously. In other words, multiple measures are organically executed to ensure final quality, such as modulating the spindle speed in real-time during processing to suppress chatter, and providing intermediate inspection guidelines to the operator after processing is complete.

[0205] This comprehensive embodiment clearly demonstrates the core technical features and effects of the invention, such as multi-perspective analysis, organic interaction, preemptive and preventive control, and human-machine collaboration. By closely exchanging data and interacting with each other, each component and multiple processes realize comprehensive and intelligent process control that was impossible with fragmentary information alone.

[0206] Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. Various modifications are possible by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention. Explanation of the symbols

[0207] Data collection unit (100) Data processing module (110) State prediction module (120) Processing parameter generation module (130) Device control unit (200)

Claims

Claim 1 An intelligent metalworking control device comprising: a data collection unit for collecting metalworking process data including a plurality of sensors; a data processing module for processing the collected process data; a state prediction module for determining a plurality of process states, including abnormal tool conditions, machining stability risks, cutting fluid contamination conditions, and operator-cognitive load conditions, based on the processed process data, and predicting final machining quality by synthesizing the determined plurality of process states; a machining parameter generation module for generating control commands to modify real-time machining conditions according to the determined process states and predicted machining quality; and a device control unit for transmitting the generated control commands to a machine tool; wherein the state prediction module comprises the steps of: calculating tool micro-profile drift by comparing shape data of a tool cutting edge acquired from a 3D laser scanner of the data collection unit with initial shape data stored in a database; and calculating a tool batch-lot freshness index by querying manufacturing lot information and history information of the tool from a database. The method for determining a tool abnormality includes the step of calculating digital log synchronization consistency by measuring timestamp errors between multiple sensors within the data collection unit; wherein, if the digital log synchronization consistency exceeds a preset first threshold, the method determines a 'warning' state requiring tool replacement as the top priority regardless of the state of the tool micro-profile drift and the tool placement-lot freshness index, thereby preventing misjudgment due to reduced reliability of the measurement data itself; and the state prediction module includes the step of calculating machining chatter phase synchronization by performing cross-correlation analysis between vibration data received from first and second acceleration sensors, respectively attached to the spindle housing and the workpiece fixing jig, which are included in the data collection unit.An intelligent metalworking control device comprising: a step of calculating power distortion-phase skew by receiving voltage and current waveforms supplied to a spindle motor from a power quality analyzer included in the data acquisition unit and analyzing the phase difference thereof; and a step of calculating a process noise spectrum anomaly index by calculating the difference between the frequency spectrum of process noise collected from a microphone of the data acquisition unit and a previously stored steady-state spectrum; wherein the state prediction module determines an 'initial unstable state' by judging it as an early sign that may develop into a serious problem, even if the processing chatter phase synchronization is below the 'stable' state threshold, when the calculated power distortion-phase skew exceeds a second threshold and simultaneously the calculated process noise spectrum anomaly index exceeds a third threshold; and wherein the processing parameter generation module generates a spindle speed modulation command that continuously changes the spindle rotation speed within a certain range according to the result of determining the 'initial unstable state', thereby preemptively avoiding conditions for the occurrence of self-excited vibration before mechanical chatter occurs through a combination of electrical instability and acoustic anomaly signals. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 An intelligent metalworking control device according to claim 1, wherein the data processing module matches the line number (N-code) of a currently running NC program received from a computer numerical control (CNC) device to all data received from the data collection unit, assigns a high-precision timestamp based on International Standard Time (UTC), and stores it in a time-series database; and the device control unit converts the control command generated from the machining parameter generation module into a data packet supporting the MTConnect or OPC-UA (Open Platform Communications Unified Architecture) standard communication protocol and transmits it to a machine tool, thereby ensuring compatibility with heterogeneous machine tools and ensuring that all control operations are accurately linked to specific machining steps and time information. Claim 6 An intelligent metal processing control device according to claim 1, wherein the state prediction module receives feedback on past actual processing results and automatically adjusts threshold values ​​for determining a 'caution' or 'warning' state through a machine learning-based reinforcement learning algorithm, and, if a defect occurs in the final quality inspection despite processing in a 'caution' state, lowers the threshold value of the judgment criterion that caused the 'caution' state in a conservative direction during the next processing, thereby improving the judgment accuracy through actual production results.

Images