Method for deriving and correcting correlation between virtual cutting force data and physical machining load data in CNC machining, and apparatus for implementing said method

Abstract

The present invention relates to a method and apparatus for deriving and correcting a correlation between cutting force simulation data and actual machining load data, and comprises: a) a step of collecting machining position data from a CNC apparatus; b) a physical data collection step of collecting physical data via an IoT sensor; c) a virtual data acquisition step of generating virtual data via simulation software; d) a data mapping step of mapping the physical data and the virtual data on the basis of the same machining position; and e) a data correction step of deriving a correlation between the physical data and the virtual data on the basis of data about the mapping, and correcting data on the basis of same.

Description

A method for deriving and correcting the correlation between virtual cutting force data and physical machining load data in CNC machining, and a device for implementing the method.

[0001] The present invention relates to a method for deriving and correcting the correlation between cutting force simulation data and physical data of machining load, and a system for implementing the method. Specifically, it provides a correction method and an apparatus that enable the diagnosis of tool wear conditions, etc., in a simulation by reflecting physical data into virtual data generated by the simulation.

[0002] In recent manufacturing technology, multi-product, small-batch production methods are continuously expanding to meet the diverse needs of users. From the perspective of production companies, this type of production strongly demands flexibility and intelligence in production technology; consequently, the utilization and application of manufacturing data are attracting attention as key technologies for such intelligent and flexible production. To facilitate the practical application of manufacturing data, technological development is required to implement Cyber-Physical Systems (CPS) and application solutions. This trend is similar in the field of CNC machining (including cutting and machining), which accounts for one of the most significant manufacturing processes.

[0003] The digital transformation of manufacturing processes utilizing digital data is expected to contribute to increased productivity and quality improvement by being applied to the optimal layout of manufacturing lines for multi-product, small-batch production and the optimization of process execution times for each piece of equipment. Related fields include Digital Twin, Virtual Manufacturing, and the Metaverse, and the technology of the present invention can also be applied to these fields.

[0004] Meanwhile, building a digital twin for CNC machining requires various types of data related to the process. In reality, the difficulty of obtaining this data varies depending on the specific item. For example, calculating cutting force and material removal volume through simulations can be easily done using cutting mechanics. However, in actual machining, cutting force requires installing expensive sensors in a highly stable environment, and there are no sensors available to detect material removal volumes, such as chips, in real time.

[0005] In addition, as shown in [Figure 1], the tool wears down as machining progresses, but it is difficult to predict the state and characteristics of tool wear in simulation. Although the machining load increases and variability rises due to tool wear during actual machining, there is no change in cutting force in the simulation. The reason it is difficult to reflect tool wear in the simulation is, first, that the precise mechanism of tool wear has not been accurately established to date, making theoretical calculations difficult, and second, that the brittleness of the carbide tool material produced using powder metallurgy manufacturing technology is a major cause.

[0006] Moreover, while the accuracy and speed of digital twins can be improved by directly utilizing physical data collected through edge devices and IoT gates, it has been difficult to ensure effectiveness in manufacturing methods requiring flexible responses, such as high-mix, low-batch production, because users spend a significant amount of time inputting parameters in digital twins for CNC equipment and machining processes.

[0007] Accordingly, the inventor of the present invention conceived a method to enable the direct use of physical data collected through edge devices and IoT gates in a digital twin.

[0008] The present invention aims to provide a method for finding the correlation between physical data and virtual data and a correction method based on the correlation in order to implement a digital twin usable in an actual manufacturing site.

[0009] The present invention aims to enhance the usability and scalability of a digital twin by utilizing actual data and corrected virtual data together under the precondition that the generated data from the simulation becomes closer to the actual manufacturing process state data and sufficient reliability is ensured.

[0010] To solve the above problem, the present invention,

[0011] It includes a process of mapping physical data and virtual data to the same processing location, comparing and analyzing the two types of data, deriving correlations, and generating correction formulas.

[0012] The present invention presents a method for utilizing virtual data in cases where its use is advantageous for constructing a digital twin in CNC machining, and for using virtual data by making it approximate actual data through correlation and correction.

[0013] In conventional technology, the precise mechanism of tool wear has not been accurately established to date, making theoretical calculations difficult. Furthermore, due to the high brittleness of carbide tool materials produced by powder metallurgy, it is difficult to reflect tool wear in simulations. However, the present invention enables the diagnosis of tool wear conditions in simulations by reflecting actual processing load data in the simulation data.

[0014] In addition, since actual physical data can be directly input into the digital twin and utilized during CNC machining through edge devices, it can be very advantageous for the automation and efficiency of data processing.

[0015] Figure 1 is a conceptual diagram showing the difference between actual machining load data and simulation data in CNC machining.

[0016] FIG. 2 is a flowchart for carrying out the present invention.

[0017] Figure 3 illustrates the characteristics of each type of data handled in the present invention.

[0018] FIG. 4 illustrates a mapping method between data for implementing the present invention.

[0019] Figure 5 is an explanatory diagram illustrating a correction method that considers cutting force in the mapping between machining load data and cutting force data, and a machining load correction method due to tool wear.

[0020] FIG. 6 illustrates a system configuration diagram for implementing the present invention.

[0021]

[0022] The details regarding the implementation of the present invention are described in detail below through the attached drawings and description.

[0023] Figure 1 is a diagram illustrating the relationship between cutting force simulation data and machining load IoT data corresponded by machining section in general CNC machining. As shown in Figure 1, the cutting force simulation data can be divided into machining section 1, which includes the initial stage of machining; machining section 2, where tool wear progresses; machining section 3, where tool wear worsens; and machining section 4, where the tool breaks. The cutting force simulation data for each section reflects changes in cutting resistance according to tool condition and machining conditions, and allows for the visual identification of specific phenomena in the process. Meanwhile, the machining load IoT data represents current-based load changes measured by actual sensors. By comparing and contrasting changes in the machining section, such as tool wear and breakage, with the cutting force simulation data, the IoT data can detect abnormal conditions occurring in the actual process.

[0024]

[0025] For example, in machining section 2, the IoT data shows a tendency to gradually increase, which coincides with the progression of tool wear. In addition, in machining section 3, a pattern of rapid increase in load is observed, confirming that this is a section where tool wear worsens. Finally, in machining section 4, the IoT data fluctuates rapidly or decreases, suggesting tool breakage.

[0026] Accordingly, the present invention enables the complementary utilization of simulation and actual measurement data by correlating cutting force simulation data with IoT-based machining load data. As a result, it allows for the precise analysis of process conditions that are difficult to ascertain from single data alone, and can significantly contribute to tool condition monitoring, anomaly detection, and ensuring machining stability.

[0027]

[0028] Specific details regarding the implementation will be explained with reference to [Fig. 2]. First, Step 1 of [Fig. 2] is a step of inputting source data for CNC machining digital twin. The source data includes CNC data (1-A) containing machining position data, etc., IoT data (1-B) containing actual machining load data, etc., and simulation data (1-C) containing cutting force data, etc.

[0029]

[0030] Before describing FIG. 2, FIG. 3 shows a comparison of the characteristics of processing data according to an embodiment of the present invention. As shown in FIG. 3, the processing data is broadly divided into processing position data (1-A), processing load data (1-B), and cutting force data (1-C).

[0031]

[0032] The machining position data (1-A) is acquired directly through communication with the CNC equipment, and the unit is mm, with the domain defined as a position series. Although the data is useful for calculating the feed average of the machining process, it is difficult to clearly separate individual machining sections because the acquisition interval is not constant and is recorded at unequal intervals. In addition, there are limitations in analyzing process characteristics as it does not include geometric shape information.

[0033]

[0034] The machining load data (1-B) is collected through an IoT-based current sensor, with a unit of A and a time series domain. The data is acquired at regular time intervals according to the sensor sampling period, and the trend of load fluctuation can be identified through a moving average. However, it is difficult to clearly distinguish the segmental characteristics of each tool, and, like machining position data, geometric information is not included.

[0035]

[0036] The cutting force data (1-C) is calculated through cutting simulation, with the unit being N and the domain being the length of the machining process. The cutting force data reflects the cutting resistance for each process and allows for the direct identification of changes based on the division of the machining section and the cutting conditions of the tool. In particular, since the cutting force data includes geometric information of the machining process, it can be effectively utilized for the division of the machining section and process analysis, unlike other data.

[0037]

[0038] Accordingly, the present invention enables the clear division of process sections and precise analysis of machining status—which cannot be provided by a single data source—by complementarily combining machining position data, machining load data, and cutting force data having different characteristics as described above. This provides a key technical effect that can contribute to future process optimization, defect prediction, and productivity improvement.

[0039]

[0040] The characteristics of the three types of data mentioned above are explained through [Figure 3]. The main characteristics of the data for each type are as follows:

[0041] CNC data (1-A) is data acquired while the CNC machine is actually machining with a tool, and it lacks geometric shape information and has the characteristic that the interval of the machining position data is non-uniform in terms of time and position. This is because, when machining metal materials with high physical resistance, the CNC prioritizes allocating most of its resources to the machining operation and handles communication with the outside as a lower priority.

[0042] IoT data (1-B) is data showing the processing position, with a unit of mm. Due to the influence of CNC communication, the processing position acquisition time fluctuates dynamically, resulting in unequal intervals. It is uniform time-series processing load data for calculating a moving average and is acquired at a very fast detection speed. Since the sensor detects the physical processing state in a time series, information such as processing shape is absent.

[0043] Simulation data (1-C) is an output that reflects the product shape, material shape, tool shape and physical characteristics, and provides geometric shape information for the processing section and processing position and cutting force data at uniform intervals equal to the chip thickness.

[0044] The above three types of data are segmented and digitized according to their respective characteristics. In addition, data items that cannot be obtained from individual types of data can be generated by combining the above three types of data.

[0045]

[0046] Step 2 of [Fig. 2] is the step of classifying and dividing the machining section using simulation data. As explained above, the simulation data includes machining positions and cutting forces based on the geometric shape within the machining section. Although the mechanism regarding tool wear in cutting processes has not been precisely established, it is known that changes in cutting force have a significant impact on the tool wear state. When the geometric shape of the machining section changes, the material removal amount and cutting force change. For example, if the machining section changes from a straight line shape to an inner circle shape, the cutting force increases in the simulation. In other words, since the actual machining load and cutting force basically show similar trends, the fundamental trend of the machining load caused by the cutting force can be identified by dividing the machining section according to geometric characteristics. Since the change in material removal amount according to the geometric shape is nonlinear, the cutting force also changes nonlinearly. Similarly, the machining load is also nonlinear because it is controlled by the spindle motor torque curve, but the difference from the cutting force changes only within an allowable range. For example, while the cutting force can yield a very large value through simulation, the machining load may have a smaller value than the simulation due to the control tolerance range based on the limits of the spindle motor. For the above detailed sections, the mapping method is valid, but the cutting force and machining load must be processed using a separate correction formula. The specific method is presented in Step 4 of [Fig. 2].

[0047] Step 3 of [Fig. 2] is to perform mapping to link three types of data (1-A, 1-B, 1-C) in the processing section divided by the geometric shape. The specific mapping method is explained together with [Fig. 4]. First, the 'processing position data via CNC communication (1-A)' collected from the actual object and the 'IoT data regarding processing load (1-B)' are synchronized. Then, as seen in '① Processing Position Mapping', the processing position P (1-A) collected from the CNC equipment is mapped to 1-B and called P' (1-A).

[0048] Next, P(1-B) is interpolated according to the geometric shape within the machining section of 1-C to form a machining path domain (i.e., a series of machining positions along the machining path) to obtain the next P'(1-C). Then, as seen in '② Machining Position Mapping', P'(1-C) can be obtained by mapping the machining position P(1-C) in the simulation to 1-B, in the same way as projecting it onto the machining path domain obtained by interpolating P'(1-B).

[0049]

[0050] Figure 4 will be explained in detail. Figure 4 is a diagram illustrating the process of mapping and aligning physical data and digital data by processing section unit according to an embodiment of the present invention. At the top of the diagram, processing position data (1-A) obtained through CNC communication is displayed, and the position is defined by section according to the classification of processing sections based on geometric shape. This processing position data (P(1-A)) is converted into normalized processing position data (P′(1-A)) through a mapping process (① processing position mapping step).

[0051]

[0052] Processing load data (1-B) collected through IoT sensors is shown in the middle,

[0053] Since this is collected based on the time axis, it does not directly correspond to individual processing sections. Therefore, the processing load data is mapped to each section based on the normalized processing position data (P′(1-A)).

[0054]

[0055] Simulation data (1-C) for cutting force is shown at the bottom, which is expressed as a predicted cutting force value reflecting the machining process conditions. The data (P(1-C)) is then mapped onto the same coordinate axis as the machining position and machining load data, and aligned into normalized cutting force data (P′(1-C)) (② Machining position mapping step).

[0056]

[0057] Accordingly, as illustrated in FIG. 4, the present invention can implement a digital twin environment in which real data and digital data can be mutually compared and verified by mapping and aligning ① CNC machining position data, ② IoT-based machining load data, and ③ simulation cutting force data into a single integrated machining section unit.

[0058]

[0059]

[0060] After performing data mapping as described above, Step 4 of [Fig. 2] is the step of performing data correction. This will be explained specifically through [Fig. 5].

[0061] In the following description of the invention, ‘IoT data including actual processing load data, etc.’ is referred to as ‘processing load data,’ and ‘simulation data including cutting force data, etc.’ is referred to as ‘cutting force data.’

[0062] First, the machining load data in 1-B and the cutting force data in 1-C are corrected. Step 3 of [Fig. 2] above is a step of mapping the machining load data in 1-B and the cutting force data in 1-C to each machining position in the machining path domain. The cutting force data is a cutting force based on cutting physics determined solely by the material shape, material properties, tool shape, tool specifications, etc. In this state, there are machining sections where a large difference occurs between the machining load data and the cutting force data; this occurs when the cutting force is very high but the actual machining load data is significantly smaller compared to the cutting force. This may be due to the use of a material different from the process plan or a defective material, or conversely, if the cutting force is very low but the actual machining load is high, it is attributed to a change in material or a change in machining position. By step 4-A, the corresponding section is considered to have material abnormalities or defects and is classified as a peculiarity, and further corrections in steps 4-B and 4-C are not performed.

[0063] Meanwhile, variations such as tool wear that fluctuates during machining and variations caused by frictional heat between the tool and the material cannot be reflected in the simulation. In other words, the actual machining load change is a mixture of variable factors, such as the cutting force resulting from material removal and tool wear occurring during machining. 4-B is a step for correcting the machining load by considering the cutting force; the actual machining load is corrected by considering the magnitude of the cutting force in P(1-C), and this is denoted as Lf. A correlation equation is used between the magnitude of the actual machining load in P(1-B) derived from the pre-tool wear section and the magnitude of the cutting force in P(1-C). The above correlation equation is obtained using a non-linear regression equation, and the magnitude of the cutting force is corrected by the machining load magnitude Lf calculated by the correlation equation established by the above non-linear regression equation, and this correction equation is as follows.

[0064] The formula for deriving the correction factor F for cutting force and machining load in the tool condition excluding wear is as follows.

[0065]

[0066] Here, F is the correction factor, LSM is the least squares method, f(1-B) is the actual processing load data, and g(1-C) is the cutting force data.

[0067]

[0068] Here, Lf is a value corrected from actual processing load data by reflecting only the cutting force, and F and g(1-C) are as above.

[0069]

[0070] As shown in [Fig. 5], if the above Lf is excluded from the actual machining load, Lw remains. Lw is the machining load that occurs due to fluctuations caused by the intensification of tool wear during the machining process mentioned earlier. 4-C in [Fig. 2] is a correction of the machining load by considering the actual machining load data while including tool wear. Furthermore, if the magnitude of Lw, Lw variance, and standard deviation deviate from the allowable range, it can be diagnosed that tool wear has progressed to the point where it can adversely affect machining quality.

[0071]

[0072] Here, Lw is a value corrected for the actual machining load considering tool wear, and Lf and f(1-B) are as mentioned above.

[0073] Through the above equations, actual machining load data and cutting force data can be corrected and utilized as approximate values, and even in cases where actual machining load data is unavailable, simulation data can be corrected and utilized in digital twins or virtual manufacturing solutions.

[0074] In addition, by configuring a device and system that derives and corrects the correlation between cutting force data and actual processing load data according to the functional part and module of [Fig. 6], it can be applied to digital twins, virtual manufacturing solutions, digital transformation manufacturing solutions, etc. To explain Fig. 6 in detail, Fig. 6 shows a block diagram of a data processing device for CNC machining digital twin according to an embodiment of the present invention.

[0075]

[0076] As illustrated in the drawing, the device is largely composed of ① a source data input unit, ② a processing section division unit, ③ a data mapping unit, and ④ a data correction unit.

[0077] First, the source data input section for the CNC machining digital twin includes a CNC data module (machining position, etc.), an IoT data module (machining load, etc.), and a simulation data module (cutting force, etc.), and simultaneously receives data collected during actual machining and simulation prediction data.

[0078] Next, the processing unit for classifying machining sections using simulator data subdivides the machining sections using geometric information based on simulation data, which is subsequently used as a standard for mapping. The machining position series mapping unit between dual data for the sections

[0079] By mapping CNC data and machining load data based on the machining position, or mapping machining load data and cutting force data, physical data and digital data are aligned in the same segment unit.

[0080]

[0081] Finally, the processing unit that corrects the actual machining load and cutting force data for the above section corrects the data by considering factors such as the torque curve, cutting force, and tool wear. Specifically, it includes a machining load data classification module considering the torque curve, a machining load correction module considering the cutting force, and a machining load correction module considering tool wear.

[0082] Accordingly, as illustrated in FIG. 6, the present invention can realize a digital twin environment for CNC machining by integrally processing real data (CNC, IoT) and digital data (simulation) to provide a high-precision dataset that is mapped and corrected in units of machining sections.

[0083]

[0084]

[0085] The present invention provides a method for utilizing virtual data in cases where it is advantageous to use virtual data to build a digital twin in CNC machining, while using it in a way that approximates actual data through correlation and correction. Since actual physical data can be directly input into the digital twin during CNC machining via an edge device, it is highly advantageous for the automation and efficiency of data processing, making it industrially useful.

Claims

1. A CNC machining process data correction method that maps physical machining load data and virtual cutting force data at the same machining position to derive a correlation and performs correction based on the correlation, a) A step of collecting machining position data from a CNC device; b) A physical data collection step for collecting physical data through IoT sensors; c) A virtual data acquisition step that generates virtual data through simulation software; d) a data mapping step for mapping the physical data and virtual data based on the same processing location; and e) a data correction step of deriving the correlation between the physical data and the virtual data based on the mapping data and correcting the data based thereon; a method for deriving and correcting the correlation between virtual cutting force data and physical machining load data in CNC machining.

2. In Paragraph 1, The above physical data is actual processing load data, and the above virtual data is cutting force data generated through simulation software, A method for deriving and correcting the correlation between virtual cutting force data and physical machining load data in CNC machining, characterized by the above correlation analysis step linking physical data and virtual data at the same machining position of the CNC machining path and applying an interpolation method that considers the geometric shape for each machining position.

3. In Paragraph 2, A method for deriving and correcting the correlation between virtual cutting force data and physical machining load data in CNC machining, characterized in that the equation for deriving the above correlation is obtained as a non-linear regression equation, and the magnitude of the cutting force is corrected by the magnitude of the machining load Lf calculated by the correlation equation (Equation 1) and the correction coefficient derivation equation (Equation 2) established by the above non-linear regression equation. (Equation 1) (Here, Lf is the value obtained by correcting the actual machining load data by reflecting only the cutting force, F is the correction factor, and g(1-C) refers to the cutting force data.) The formula for deriving the correction factor F of the above cutting force and machining load is as follows. (Equation 2) (Here, F is the correction factor, LSM is the least squares method, and f(1-B) is the actual processing load data.) 4. In Paragraph 3, A method for deriving and correcting the correlation between virtual cutting force data and physical machining load data in CNC machining, characterized in that Lw, a machining load generated by fluctuations such as intensified tool wear occurring during machining, is calculated by excluding the above Lf from the actual machining load and correcting the actual machining load data according to the following relationship (Equation 3). (Equation 3) (Here, Lw is the value corrected for the actual machining load due to tool wear.) 5. An apparatus for mapping physical machining load data and virtual cutting force data, deriving and correcting correlations, and generating and utilizing the corrected data for constructing a digital twin in a CNC machining process, a) CNC equipment that collects machining position data during cutting; b) IoT data input module for collecting physical data; c) Simulation data generation module that generates virtual data; d) A data mapping module that maps collected physical data and virtual data based on the same processing location; e) A CNC machining data correlation derivation and correction device comprising a data correction module that analyzes mapped data to derive correlations and performs data correction based on the correlations.

Images