Device and method for compensating data loss in industrial machine

The data loss completion device for industrial machinery addresses frequent data loss in servo waveform data by generating estimated waveform data to supplement defects, enhancing reliability and prediction capabilities while minimizing data acquisition points and network load.

WO2024247258A1PCT designated stage expired Publication Date: 2024-12-05FANUC LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2023/020661
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-06-02
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Industrial machinery experiences frequent data loss in servo waveform data, which affects the reliability of data analysis and prediction of failures and wear, as existing methods require multiple data sets and are not efficient in filling gaps without compromising bias.

Method used

A data loss completion device and method that calculate feature amounts from both actual and simulated servo waveform data, optimize execution parameters, and generate estimated waveform data to supplement defects, allowing for efficient data completion without requiring multiple data sets.

Benefits of technology

Enables reliable data analysis and prediction of failures and wear by minimizing data acquisition points, reducing network load, and increasing the amount of training data, even with small amounts of actual data, while maintaining accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023020661_05122024_PF_FP_ABST
    Figure JP2023020661_05122024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention provides a device and method for compensating a data loss in an industrial machine that is capable of compensating a loss once a certain amount of data is acquired. The device for compensating a data loss comprises: a matching unit that performs matching between first servo waveform data, which is actually measured, and an NC program, and obtains a correspondence relationship between the first servo waveform data and the NC program; a waveform generation unit that generates second servo waveform data by simulation using the NC program; a correspondence relationship analysis unit that obtains, on the basis of the operation information about the NC program and the number of executions of the NC program and the first servo waveform data, whether there is a loss in the first servo waveform data after the number of executions, and a correspondence relationship between the number of executions and the first servo waveform data; and a data loss compensating unit that compensates the loss by using the second servo waveform data when there is the loss in the first servo waveform data.
Need to check novelty before this filing date? Find Prior Art

Description

Industrial machine data loss compensation device and data loss compensation method

[0001] The present disclosure relates to a data loss compensation device and a data loss compensation method for industrial machinery, and in particular to a data loss compensation device and a data loss compensation method that compensates for losses in actually measured servo waveform data (time series data) of industrial machinery by using servo waveform data (time series data) generated by simulation.

[0002] Servo waveform data is the source data from which the state of motors, machines, and processing can be inferred. It can also be used to detect and predict failures and wear, so it is used in various places. When handling servo waveform data, data loss is a frequent phenomenon. A technology for compensating for data loss when it occurs is described, for example, in Patent Document 1. Technologies related to the present disclosure are described, for example, in Patent Document 2 and Patent Document 3.

[0003] Patent Literature 1 describes a missing value imputation device that can impute missing values ​​while suppressing bias within a specified time. Specifically, Patent Literature 1 describes that the missing value imputation device that imputes missing values ​​of collected data includes: an imputation processing unit that measures a required time for imputation processing by simulation and associates the required time with each multiplexing number of data multiplexed by multiple imputation, a multiplexing number acquisition unit that, when an allowable time for imputation processing is specified, acquires a multiplexing number at which the imputation processing will be completed within the allowable time based on the associated required time, and a calculation method determination unit that determines a calculation method for the imputation processing according to the acquired multiplexing number.

[0004] Patent Document 2 describes an equipment monitoring and control device that is highly accurate and reliable even when parameters of a drive system or mechanical system model fluctuate. Specifically, Patent Document 2 describes an equipment monitoring and control device that monitors or controls a processing machine using a model that represents the drive system or mechanical system of the processing machine, has a database that stores correction values ​​of model parameters that correspond to a plurality of items related to drive commands, and corrects the parameters using the database to monitor or control the processing machine.

[0005] Patent Document 3 describes a numerical control device that can detect cutting abnormalities with high accuracy. Specifically, Patent Document 3 describes that the numerical control device is a numerical control device with a load torque measurement function for a servo motor or a spindle motor, and includes means for recognizing the degree of tool wear, storage means for storing a plurality of limit values ​​corresponding to the degree of tool wear for the load torque of each motor, interpolation means for determining the limit value corresponding to the degree of tool wear from the limit values ​​stored in the storage means by interpolating, and load torque comparison means for comparing the limit value determined by the interpolation means with the load torque measured by the measurement function, thereby monitoring abnormalities in the load torque.

[0006] JP 2020-177560 A JP 2000-33532 A JP 2004-252839 A

[0007] When attempting to analyze data containing missing data, the amount of data is reduced by the amount of missing data, reducing the reliability of the study. When imputing missing data, it is desirable to be able to imputate missing data by acquiring a certain amount of data once, rather than preparing multiple datasets with filled missing data, as in multiplex imputation.

[0008] A first representative aspect of the present disclosure includes a feature calculation unit that calculates a first feature including a statistical quantity of torque included in first servo waveform data based on actually measured first servo waveform data, and calculates a second feature including a statistical quantity of torque included in the second servo waveform data based on second servo waveform data; a correspondence relationship analysis unit that determines, based on the number of executions of an NC program that drives an industrial machine and the first servo waveform data, whether or not the first servo waveform data has any missing data in the number of executions and a first correspondence relationship between the number of executions and the first servo waveform data; a parameter optimization unit that optimizes execution parameters including at least a cutting coefficient so that the first feature amount and the second feature amount coincide with each other, and determines a second correspondence relationship between the execution parameters and the first servo waveform data; a waveform generation unit that generates the second servo waveform data by a simulation using the NC program and the execution parameters, and determines a third correspondence relationship between the second servo waveform data and the execution parameters; and a data defect complementing unit that, when there is a defect in the first servo waveform data, complements the defect by using the second servo waveform data.

[0009] A second representative aspect of the present disclosure is a computer-implemented method for computing a first feature amount including a torque statistic included in first servo waveform data based on actually measured first servo waveform data, and computing a second feature amount including a torque statistic included in the second servo waveform data based on second servo waveform data; a process for determining, based on the number of executions of an NC program for driving an industrial machine and the first servo waveform data, whether or not there is a defect in the first servo waveform data for the number of executions and a first correspondence relationship between the number of executions and the first servo waveform data; a process for optimizing execution parameters including at least a cutting coefficient so that the first feature amount and the second feature amount coincide with each other, and determining a second correspondence relationship between the execution parameters and the first servo waveform data; a process for generating the second servo waveform data by a simulation using the NC program and the execution parameters, and determining a third correspondence relationship between the second servo waveform data and the execution parameters; and a process for complementing the defect using the second servo waveform data if there is a defect in the first servo waveform data. This is a method for completing missing data in industrial machinery.

[0010] 15 is a block diagram showing the overall configuration of a data loss complementing device that complements data loss in industrial machinery according to an embodiment of the present disclosure. FIG. 15 is a diagram showing a process of calculating an average value of torque data in a certain section as an actual waveform feature value and calculating a transition graph of the actual waveform feature value. FIG. 15 is a diagram showing an example of operation information. FIG. 15 is a diagram showing count information, operation information, an actual waveform, and an estimated waveform. FIG. 15 is a diagram showing an example of an actual waveform. FIG. 15 is a diagram showing an example of complementing a partial loss of an actual waveform in an NC program with an estimated waveform. FIG. 15 is a characteristic diagram showing time changes of an actual measured torque value and an estimated torque value calculated by simulation using a cutting coefficient. FIG. 15 is a characteristic diagram showing the relationship between the average value of the estimated torque values ​​and the cutting coefficient Ktc. FIG. 15 is a characteristic diagram showing the relationship between the average value of the estimated torque values ​​and the cutting coefficient Kte. FIG. 15 is a block diagram showing the configuration of a model representing a controller. FIG. 15 is a block diagram showing the configuration of a spindle model. FIG. 15 is a diagram showing an example of servo waveform data, which is an actual waveform. FIG. 15 is a diagram showing how matching is performed in the order of same tool section, same speed section, or cutting feed signal section.

[0011] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings.

[0012] 1 is a block diagram showing the overall configuration of a data loss correction device for correcting data loss in industrial machinery according to an embodiment of the present disclosure. The data loss correction device 10 includes a matching unit 101, a feature calculation unit 102, a correspondence relationship analysis unit 103, a state estimation unit 104, a data loss correction unit 105, a parameter optimization unit 106, a waveform generation unit 107, a storage unit 108, and a waveform correlation unit 109. The data loss correction device 10 may be provided as part of the industrial machinery or separately from the industrial machinery. Data loss is also referred to as missing data.

[0013] The industrial machines for which the data loss compensation device 10 compensates for data loss are, for example, machine tools and cutting robots.

[0014] The feature calculation unit 102, the correspondence analysis unit 103, the data loss complementation unit 105, the parameter optimization unit 106, and the waveform generation unit 107 constitute a component (hereinafter referred to as a "data loss complementation component") for realizing a function of complementing missing data using estimated servo waveform data (hereinafter referred to as an estimated waveform) generated by the waveform generation unit 107 when there is a loss in actually measured servo waveform data (hereinafter referred to as an actual waveform). The actually measured servo waveform data that becomes the actual waveform becomes first servo waveform data, and the servo waveform data that becomes the estimated waveform becomes second servo waveform data. The waveform generation unit 107 generates the estimated waveform by simulation using an NC program.

[0015] Other than the missing data complementation configuration unit, the matching unit 101, state estimation unit 104, storage unit 108, and waveform correlation unit 109 may be provided as needed. The waveform generation unit 107 may utilize one or more external tools, and if the waveforms generated are appropriate and the time cost for generation is sufficiently fast, it may be possible to replace the waveform generation unit 107 in the future.

[0016] First, the missing data compensation component in the missing data compensation device 10 will be described, and then each component other than the missing data compensation component will be described. Below, each component included in the missing data compensation component will be described.

[0017] (Feature Calculation Unit 102) The feature calculation unit 102 calculates feature amounts (hereinafter referred to as real waveform feature amounts) for a certain interval using the real waveform, and outputs the calculated feature amounts to the storage unit 108. It is desirable to use an actual waveform that has been confirmed to match with the NC program by the matching unit 101, which will be described later. The certain interval is, for example, an interval of change in the movement amount of a position command, and the real waveform feature amount for the certain interval is, for example, at least one of statistics of torque data such as the mean, variance, skewness, kurtosis, and RMS (root mean square) of torque data (for example, a spindle torque command) for the certain interval.

[0018] For example, as shown in Fig. 2, the feature calculation unit 102 calculates, as the actual waveform feature, the average value of torque data in a fixed interval R1 (shown as "interval R1" in Fig. 2) and a fixed interval R2 (shown as "interval R2" in Fig. 2) that are divided based on the change in the movement amount of the X-axis position command. Then, the actual waveform feature values ​​in the fixed intervals R1 and R2 are collected for each program execution count, and a transition graph of the actual waveform feature values ​​in the fixed intervals R1 and R2 is calculated. Fig. 2 is a diagram showing the process of calculating the average value of torque data in the fixed intervals as the actual waveform feature value and calculating the transition graph of the actual waveform feature value.

[0019] Furthermore, the feature calculation unit 102 calculates a feature for a certain interval (hereinafter referred to as an estimated waveform feature) using the estimated waveform generated by the waveform generation unit 107, and outputs the calculated feature to the storage unit 108. The method for calculating the estimated waveform feature is the same as the method for calculating the real waveform feature, and the calculation can be performed by replacing the real waveform with the estimated waveform. The certain interval for calculating the estimated waveform feature corresponds to the certain interval for calculating the real waveform feature.

[0020] (Correspondence analysis unit 103) The correspondence analysis unit 103 determines whether there is a defect in the actual waveform for each number of times the NC program is executed (including whether it is an NC program unit or a part of the NC program) based on the operation information, count information, and actual waveform, and stores the correspondence between the number of times the NC program is executed and the actual waveform in the storage unit 108. The correspondence between the number of times the program is executed and the actual waveform becomes a first correspondence.

[0021] The correspondence relationship analysis unit 103 acquires operation information, count information, and actual waveforms from the memory unit 108. Operation information is actual data on the operation of an NC program. The operation information reveals the NC program name, start time, and execution period (end time). Figure 3 is a diagram showing an example of operation information. Operation information has a longer period than servo waveform data, which is an actual waveform, and is acquired from a different port, so data loss is less likely to occur than with actual waveforms. For operations during periods when data cannot be obtained due to network outages or the stopping of data collection equipment, count information is acquired from the memory unit 108 and the NC program being executed is inferred.

[0022] The count information is the number of times the NC program has been executed (number of times the NC program has been executed). The count can be obtained if the CNC has a program count. The number of times a tool has been used does not necessarily match the program count, so it is acceptable to count it in the NC program.

[0023] Losses may occur when there is count information and an NC program is being executed, but there is no operation information and the actual waveform is missing for each NC program; when there is operation information but the actual waveform is missing for each NC program; or when there is operation information but part of the actual waveform in the NC program is missing. Figure 4 is a diagram showing count information, operation information, actual waveform, and estimated waveform. As shown in Figure 4, when there is a loss in the actual waveform for each NC program or part of the NC program, the data loss recovery unit 105, which will be described later, recovers the loss with the estimated waveform.

[0024] (Data missing complement unit 105) The data missing complement unit 105 acquires from the storage unit 108 whether there is a missing portion in the actual waveform for each NC program execution count (including whether it is an NC program unit or a part of the NC program). The data missing complement unit 105 acquires from the storage unit 108 the correspondence relationship between the estimated waveform and the NC program (fourth correspondence relationship) generated by the waveform generation unit 107. The data missing complement unit 105 also acquires the correspondence relationship between the actual waveform and the estimated waveform (seventh correspondence relationship) generated by the waveform correlation unit 109. The data missing complement unit 105 also acquires the actual waveform and the estimated waveform from the storage unit 108.

[0025] When a portion of an NC program contains a missing actual waveform, the missing data complement unit 105 complements the missing portion of the actual waveform using an estimated waveform. When a missing actual waveform (complete missing) exists for each NC program, the missing actual waveform is complemented for each NC program using an estimated waveform. When no operation information is available, the missing actual waveform is complemented for each NC program using an estimated waveform. The missing data complement unit 105 stores the complemented actual waveform in the memory unit 108. The complemented actual waveform can be used as source data for estimating the status of the motor, machine, and machining, and can also be used to detect and predict failures and wear. The missing data complement unit 105 complements missing actual waveforms for each NC program with an estimated waveform based on the correspondence between the estimated waveform and the NC program (fourth correspondence). When complementing missing actual waveforms for each NC program with an estimated waveform, it is necessary, for example, to determine the number of program executions for which the missing actual waveform occurred and to calculate the estimated waveform for that program execution. The actual waveform at which the program has been executed can be determined using the first correspondence relationship determined by the correspondence relationship analysis unit 103, and the estimated waveform at which the program has been executed can be determined using the fourth correspondence relationship. The missing data complement unit 105 complements a missing portion of the actual waveform in the NC program with an estimated waveform based on the correspondence relationship between the estimated waveform and the NC program (fourth correspondence relationship) and the correspondence relationship between the actual waveform and the estimated waveform (seventh correspondence relationship). When complementing a missing portion of the actual waveform in the NC program with an estimated waveform, for example, it is necessary to know the actual waveform at which the missing portion occurred, and then determine the correspondence relationships between the estimated waveform and the NC program and the correspondence relationships between the actual waveform and the estimated waveform. The actual waveform at which the program has been executed can be determined using the first correspondence relationship determined by the correspondence relationship analysis unit 103. The estimated waveform at which the program has been executed can be determined using the fourth correspondence relationship, and the estimated waveform corresponding to the missing portion of the actual waveform at that program execution can be determined using the seventh correspondence relationship.

[0026] Fig. 5 is a diagram showing an example of an actual waveform, Fig. 6 is a diagram showing an example of complementing a partial loss of an actual waveform in an NC program with an estimated waveform, and Fig. 7 is a diagram showing an example of complementing a partial loss of an actual waveform in an NC program with an estimated waveform.

[0027] By complementing the missing data, the actual waveform with the complemented missing data can be used as training data (normal value model) for creating a fault detection model. In this case, even if the amount of training data that can actually be acquired is reduced due to the missing data, the amount of data can be increased by complementing the missing data. Furthermore, by using the actual waveform with the complemented missing data, the locations from which data is acquired can be minimized. Therefore, data only needs to be acquired from the necessary locations, reducing the network load.

[0028] (Parameter optimization unit 106) The parameter optimization unit 106 optimizes the execution parameters so that the fluctuations in the actual waveform feature quantities calculated by the feature quantity calculation unit 102 coincide with the fluctuations in the estimated waveform feature quantities calculated by the feature quantity calculation unit 102. The cutting coefficients (Kte, Ktc) that serve as execution parameters used in the cutting force estimation unit 1074 are determined by executing a simulation in the parameter optimization unit 106 and stored in the storage unit 108. The parameter optimization unit 106 determines a correspondence relationship between the actual waveform and the execution parameters and stores the correspondence relationship in the storage unit 108. This correspondence relationship becomes a second correspondence relationship.

[0029] For example, the parameter optimization unit 106 optimizes the cutting coefficients (Kte, Ktc), which are execution parameters, so that the average value of the measured torque, which is the actual waveform feature, over a certain section, matches the average value of the estimated torque, which is the estimated feature obtained from the cutting force.

[0030] The first method for optimizing the parameters will be described below. The first method is a method for optimizing the cutting coefficients (Kte, Ktc) by calculating the cutting coefficients Ktc and Kte so that the estimated torque average generated by changing the cutting coefficients Ktc and Kte and the feature quantity (average) of the actual measured torque are approximately the same value.

[0031] For example, Fig. 8 shows the characteristics of the actually measured torque value and the estimated torque value calculated by simulation using the cutting coefficients (Kte, Ktc). In Fig. 8, the vertical axis represents torque value and the horizontal axis represents time. As shown in Fig. 8, the point in the torque where Ktc and Kte are adjusted is fixed. As shown in the characteristic diagram of Fig. 9, the average value of the estimated torque value and the cutting coefficient Ktc are proportional to each other, and the cutting coefficient Ktc is calculated so that the average value of the estimated torque value matches the average value of the actually measured torque value. As shown in the characteristic diagram of Fig. 10, the average value of the estimated torque value and the cutting coefficient Kte are proportional to each other, and the cutting coefficient Kte is calculated so that the average value of the estimated torque value matches the average value of the actually measured torque value.

[0032] Next, a second method for optimizing parameters is described below. The equation for the change in the cutting coefficient is optimized while determining the cutting coefficient from the change in the measured torque. The average measured torque (cutting force) changes due to wear, etc. An execution parameter (cutting coefficient) can be determined to generate an estimated waveform that matches this change. The equation for the change in the cutting coefficient is fitted to a certain equation, as the torque changes along with wear and the cutting coefficient also changes. To determine this equation, the initial cutting coefficient value and a change coefficient are required. The initial cutting coefficient value is determined by using the measured torque value immediately after tool replacement to determine the cutting coefficient value Ktc. The change equation is determined by using the equations for multiple actual measurements to determine the cutting coefficient value Ktc at each point. The changes in these cutting coefficient values ​​are calculated using an approximate equation. This becomes the equation for the change in the cutting coefficient. When tool replacement is performed several times, the changes in the cutting coefficient from multiple times are combined to determine a standardized equation for the change in the cutting coefficient (an average is also acceptable).

[0033] (Waveform generation unit 107) The waveform generation unit 107 generates servo waveform data that serves as an estimated waveform through simulation from the NC program. The waveform generation unit also generates a correspondence relationship between the estimated waveform and the execution parameters. The correspondence relationship between the estimated waveform and the execution parameters is the third correspondence relationship. The waveform generation unit 108 also generates a correspondence relationship between the estimated waveform and the NC program. The correspondence relationship between the estimated waveform and the NC program is the fourth correspondence relationship. In addition, the waveform generation unit 107 sets the execution parameters and executes a simulation to generate estimated cutting force, estimated torque, etc.

[0034] Simulation execution parameters include motor characteristics, tool information, workpiece (material) information, cutting coefficients, etc. Once the motor, tool, material, etc. are determined, most of the parameters are determined. However, the cutting coefficients (Kte, Ktc), which are execution parameters used to estimate the cutting force, are found to match the actually measured torque by running a simulation multiple times in the parameter optimization unit 106, which will be described later. Although the torque command (TCMD) can be found from the cutting coefficients, the cutting coefficients cannot be found directly from the torque command, so the cutting coefficients (Kte, Ktc) are found by running a simulation in the parameter optimization unit 106.

[0035] The following describes each unit of the waveform generating unit 107. As shown in Fig. 1 , the waveform generating unit 107 includes an NC program interpreting unit 1071, a command generating unit 1072, a feedback control unit 1073, and a cutting force estimating unit 1074. The cutting force estimating unit 1074 may be provided outside the waveform generating unit 107.

[0036] <NC program interpretation unit 1071> The NC program interpretation unit 1071 determines the travel distance, travel path, and command speed based on input information such as divided NC (Numerical Control) programs (which become machining programs) output from the storage unit 108, CNC (Computerized Numerical Control) parameters, macro variables, workpiece origin offsets, tool offset values, etc. The NC program interpretation unit 1071 interprets the NC program that defines the operations for positioning the feed axes and speed control of the spindle, breaks down the NC program into each code and value, and determines the travel distance, travel path, and command speed.

[0037] The NC program interpretation unit 1071 performs the following interpretations (A), (B), (C), and (D), for example: (A) Converts the codes M03, M04, and M05 to clockwise rotation of the spindle, counterclockwise rotation of the spindle, and stop of spindle rotation. (B) Classifies the cases into rapid traverse of the spindle to specified coordinates, cutting feed of the spindle to specified coordinates, clockwise circular interpolation, and counterclockwise circular interpolation using G00, G01, G02, and G03, and converts them into the path and travel distance of each servo axis. (C) Converts the canned cycle and tool change operations into equivalent G codes and M codes. (D) Adds the tool offset value to the travel distance.

[0038] <Command generation unit 1072> The command generation unit 1072 generates interpolation data by interpolating points on the movement path at an interpolation period based on the movement distance, movement path such as a straight line or an arc, and command speed obtained by the NC program interpretation unit 1071, generates an acceleration / deceleration profile based on the interpolation data, and further distributes the data to each control axis, thereby providing a position command value or a speed command value for each control period to the servo motor that serves as the feed axis motor and the spindle motor that serves as the spindle motor.

[0039] The command generation unit 1072 performs the following operations (E), (F), (G), (H), and (I): (E) Calculates the travel distance. (F) Creates a velocity and position profile at the tool tip that satisfies the travel distance, command velocity, and acceleration / deceleration parameters set by the machine. Considering the travel path, it is possible to add constraints, such as limiting the centripetal acceleration of the arc to a certain value. (G) Discretizes the profile for each control cycle. The velocity and position profile is generated as a function of continuous time in the previous step, and sampling from it eliminates the need to process fractions less than the control cycle. Furthermore, by handling fractions less than the control cycle for start and end times, processing is performed to prevent discontinuity between previous and next command blocks. (H) Distributes the discretized profile to commands for each axis. For straight lines, the distribution ratio of the travel distance and velocity to each feed axis is constant, but for curved lines, the distribution ratio varies depending on the position on the path. Therefore, the distribution must be calculated according to the position on the path. By distributing the position and velocity to each feed axis according to the position and travel path for each control cycle, command values ​​for each axis for each control cycle can be obtained. (I) Filtering the distributed command values ​​for each axis. Although this deviates from the original command path, it is possible to suppress sudden changes in the command values ​​and reduce mechanical vibrations and shocks.

[0040] The acceleration / deceleration and filtering processes described above are the same concept of command generation as in actual numerical control devices. Conventionally, CAM software did not handle such acceleration / deceleration and filtering processes, and simulated movement at a commanded speed. Furthermore, cutting simulations have not taken into account feed including acceleration / deceleration. However, by simulating elements such as acceleration / deceleration and filtering, it is possible to calculate required time and cycle time that are close to the actual time, which is practical when comparing with data from an actual machine.

[0041] <Feedback control unit 1073> The feedback control unit 1073 simulates feedback control that causes the rotation of the electric motors that drive the feed axis and the main axis to follow the position command value or speed command value generated by the command generation unit 1072, and generates and outputs servo waveform data that serves as an estimated waveform.

[0042] A feed axis is a mechanism that controls the relative positional relationship between the workpiece and the tool in industrial machinery. There are two types of feed axes: linear axes that perform linear translational motion, and rotary axes that perform rotational motion. This section mainly describes a model for numerically simulating the mechanism of linear axes on a computer. To simulate a feed axis, two models are required: a model that represents the feed axis, and a model of the controller that controls it.

[0043] First, we will explain the model representing the feed axis. The feed axis converts the rotational motion of the motor into translational motion through a conversion mechanism such as a ball screw. The rotation angle of the motor is represented by θ, and the displacement in the translational motion after passing through the conversion mechanism is represented by X. Furthermore, the torque applied to the motor is represented by T. In the one-inertia system model, the feed axis is considered to be a rigid body, and its motion is represented by Equation 1 (Equation 1 below). In Equation 1, J represents the moment of inertia of the drive mechanism, and D represents the viscous friction coefficient. When the rotation angle θ in Equation 1 is replaced with the rotation speed ω, Equation 1 can be expressed as Equation 2 (hereinafter Equation 2). In this case, the transfer function for the response from torque input to angular velocity output is expressed by Equation 3 (Equation 3 below) using the Laplace operator s, and Equation 3 corresponds to a first-order lag low-pass filter.

[0044] Next, a model representing the controller will be explained. The feed axis is controlled by position control with speed control as a minor loop. PI control is used within the speed control. This is expressed in a block diagram as shown in Figure 11. Figure 11 is a block diagram showing the configuration of a model representing the controller. The controller shown in Figure 11 is a controller that receives a position command X in the X-axis direction. C is converted into a rotation angle and the position command θ cmd As a result, the position feedback θ fdThe difference between the speed command ω cmd and generates a velocity feedback ω fd The controller then performs PI processing to obtain the difference between the torque command T cmd The detected angular velocity output from the feed axis model is used as the velocity feedback ω fd The detected angular velocity is integrated to obtain the detected angle, which is used as position feedback.

[0045] The transfer function representation of the configuration shown in FIG. 11 is in the continuous time domain. However, when performing a numerical simulation on a computer, it is handled in the discrete time domain, so the transfer function shown in the continuous time domain must be converted to the discrete domain. This is done according to the following rules: Blocks related to simple addition or constant multiplication are calculated by simply adding and multiplying variables. The integral term 1 / s is calculated by numerical integration. In other words, for the discrete simulation time span t and the input time series xi, the output is calculated using Equation 4 (Equation 4 below). The first-order term 1 / (1+s) corresponds to a low-pass filter. This is calculated using forward differences. For discrete simulation, it is necessary to convert the s-domain equation into a z-domain equation. For this purpose, the conversion formula shown in Equation 5 (Equation 5 below) is used. As a result, the relationship between the input X(s) and output Y(s) of the transfer function is expressed by Equation 6 (Equation 6 below) since Y(s)=X(s) / (1+s). Equation 6 provides the recurrence formula for the calculation.

[0046] Finally, we will explain the model representing the spindle. The spindle is the motor axis for rotating the tool. It is the same as the feed axis in that it drives the motor and controls the machine end, but unlike the feed axis, the machine end transmits rotational motion as is, rather than translational motion. Therefore, the spindle does not have a conversion mechanism using a ball screw. Also, unlike the feed axis, the input to the spindle is a speed command, and for internal control, there is only a speed control loop. IP control is used here. Figure 12 is a block diagram showing the configuration of the spindle model.

[0047] The feedback control unit 1073 outputs a torque command, which becomes servo waveform data created within the feedback control unit 1073 , to the cutting force estimation unit 1074 .

[0048] <Cutting force estimation unit 1074> The cutting force estimation unit 1074 estimates the cutting force of the cutting tool based on the torque command output from the feedback control unit 1073 and the characteristic values ​​stored in the memory unit 108. Here, the characteristic values ​​stored in the memory unit 108 are the quantified shape or contact length of the workpiece, etc. The cutting force is the resistance of the material against the penetration of the cutting tool and means the force required to continue cutting. Torque [N.m] can be calculated by multiplying the cutting force by the principal component [N] by the radius [m], so cutting force can be used to mean almost the same thing as torque.

[0049] The estimated cutting force can be calculated using the instantaneous cutting force model described below. The cutting edge of the cutting tool is vertically divided into infinitesimal portions. The vertical size of the infinitesimal cutting edge is defined as dz. Here, the cutting forces acting on the infinitesimal cutting edge when the cutting tool cuts the workpiece are defined as dFt, dFr, and dFa, respectively. Furthermore, h is defined as the length of the workpiece cut by the infinitesimal cutting edge in one rotation (called the cutting thickness). The cutting thickness h depends on the position z and rotation angle θ of the infinitesimal cutting edge. The instantaneous cutting model assumes that the relationship in Equation 7 (see below) holds. Here, the cutting coefficients Kte, Ktc, Kre, Krc, Kae, and Kac are coefficients determined by the physical relationship between the tool and the workpiece. In particular, the cutting coefficients Ktc, Krc, and Kac correspond to the specific cutting resistance.

[0050] If the shape of the workpiece is known in advance, it is possible to calculate the cutting force by numerically representing the shape of the workpiece in some form on a computer and simulating the cutting process at discrete time intervals. For example, a method called Z-Map considers a grid space divided in the X and Y directions, and numerically represents the shape of the workpiece by recording the height of the workpiece in each cell. The cutting tool is represented by a small plate separated in the height direction. The simulation is performed by updating the positions of the cutting tool and workpiece at short time intervals, reducing the height of the workpiece cells where the tool comes into contact, and generating cutting forces according to the thickness of the cut that occurs in those areas.

[0051] If the shape of the workpiece is not known in advance, Z-Map simulations cannot be used. However, by using torque data measured during machining, it is possible to see at what position the tool is located when torque is generated. In other words, by using the measurement data, it is possible to estimate the position of the workpiece. Furthermore, it is also possible to estimate the specific cutting resistance of the workpiece.

[0052] Here, we will explain a method for performing cutting simulation (simulation using contact length) when the specific cutting resistance is known and the workpiece shape is unknown.Since torque can be considered as the force applied in the tangential direction multiplied by the tool radius, the cutting torque dT applied to the microplate of the cutting tool is expressed by the following equation (Equation 8 below). The cutting torque dT depends on the position z and rotation angle θ of the tool microplate, but to eliminate this dependency, the average value h′ of the cutting thickness h is set as follows: Substituting the cutting thickness h in Equation 8 with the average value h' in Equation 9 and summing Equation 8 for each microplate yields Equation 10 (Equation 10 below). Here, the sum in the height direction of the tool in contact with the workpiece is set to Cz. Hereinafter, Cz will be referred to as the contact length.

[0053] In a simulation using the contact length Cz, the average chip thickness h' is calculated from the feed rate and rotation speed, which are machining conditions. If the specific cutting resistance is known, the cutting torque can be simulated by calculating the contact length Cz at each position. In practice, if measurement data is available, the contact length at each position (which serves as basic cutting data) can be calculated from Equation 10. The same applies to cutting forces in directions other than the tangential direction.

[0054] The above describes each unit of the missing data recovery configuration unit. Next, each unit other than the missing data recovery configuration unit will be described. In addition to the missing data recovery configuration unit, the missing data recovery device 10 also includes a matching unit 101, a state estimation unit 104, a storage unit 108, and a waveform correlation unit 109. Below, each unit of the matching unit 101, the state estimation unit 104, the storage unit 108, and the waveform correlation unit 109 will be described.

[0055] (Matching unit 101) The matching unit 101 matches (associates) an actual waveform acquired from an industrial machine, the actual waveform including at least a feed axis position command, a spindle torque, and a spindle speed, with an NC program that drives the industrial machine. The matching unit 101 then outputs the correspondence between the actual waveform and the NC program to the feature calculation unit 102 and stores it in the storage unit 108. The correspondence between the actual waveform and the NC program is a fifth correspondence. The matching unit 101 may acquire the actual servo waveform data directly from the industrial machine without going through the storage unit 108, or may acquire the actual servo waveform data output from the industrial machine via a CNC (Computerized Numerical Control) or a PLC (Programmable Logic Controller). The measured servo waveform data stored in the memory unit 108 may be data obtained directly from the industrial machine, or may be measured servo waveform data output from the industrial machine obtained via a CNC (Computerized Numerical Control) or a PLC (Programmable Logic Controller).

[0056] When missing parts of an actual waveform are supplemented with an estimated waveform, which is simulation data, to compare the behavior of the actual machine with the behavior in the simulation, it is necessary to match the actual waveform with the estimated waveform in order to compare the same parts. In this case, it is necessary to match the estimated waveform with the NC program, as well as the NC program with the actual waveform. This is because when using a simulation as a digital twin, it is necessary to be able to confirm that the simulation properly reproduces the behavior of the actual machine controlled by the NC program, and to be able to confirm which parts of the NC program are the behavior that should be noted when analyzing the actual machine.

[0057] When a simulation is performed to obtain an estimated waveform, the servo waveform data of the simulation can be matched with the executed lines of the NC program. Therefore, the simulation data can semantically divide the servo waveform data (estimated waveform) using the NC program and match it with the NC program. However, matching the measurement data (actual waveform) of the actual machine with the NC program is not easy, and the actual waveform may have short-term gaps due to the data acquisition process, which must be addressed.

[0058] Here, the actual waveform is divided as follows and associated with the NC program. - The actual waveform (servo waveform data) is divided into sections where each tool is used, with tool changes in the NC program as delimiters. - Next, the actual waveform is divided into sections where each tool is used in the NC program, either into constant speed sections or sections where a cutting feed signal is set. As described above, the actual waveform can be associated with the NC program in the order of tool use section, constant speed section, or cutting feed section. Note that the actual waveform can also be divided by program number or sequence number. The actual waveform can also be associated with the NC program in the order of tool use section, constant speed section, and cutting feed section.

[0059] The matching operation of the matching unit 101 is shown in Fig. 13 and Fig. 14. Fig. 13 is a diagram showing an example of servo waveform data that becomes an actual waveform. Fig. 14 is a diagram showing how matching is performed in the order of the same tool section, the same speed section, and the cutting feed signal section. Figs. 13 and 14 show both the same speed section and the cutting feed signal section.

[0060] (State estimation unit 104) The correspondence relationship analysis unit 103 determines the correspondence relationship between the number of times an NC program is executed and an actual waveform (first correspondence relationship), and the parameter optimization unit 106 determines the correspondence relationship between the number of times an NC program is executed and an actual waveform (second correspondence relationship). The waveform generation unit 107 executes a simulation and, when generating an estimated waveform, determines the correspondence between the servo waveform data of the simulation and the NC program, so that it can determine the correspondence relationship between the estimated waveform generated by the simulation and the NC program (fourth correspondence relationship) and the correspondence relationship between the estimated waveform and execution parameters (third correspondence relationship), and output them to the storage unit 108.

[0061] The state estimation unit 104 acquires from the storage unit 108 the three correspondence relationships (first to third correspondence relationships) output from the correspondence relationship analysis unit 103, the parameter optimization unit 106, and the waveform generation unit 107, and determines a correspondence relationship between the number of times the NC program is executed and an estimated waveform based on the three correspondence relationships, and outputs the determined correspondence relationship to the storage unit 108. The correspondence relationship between the number of times the NC program is executed and an estimated waveform is the sixth correspondence relationship. The first and second relationships correspond to the number of times the NC program is executed and the execution parameters, and the waveform generation unit 107 generates an estimated waveform using the execution parameters to determine the third correspondence relationship. The state estimation unit 104 determines the correspondence relationship between the number of times the NC program is executed and an estimated waveform (sixth correspondence relationship) from the correspondence between the number of times the NC program is executed and the execution parameters and the third correspondence relationship. Note that the state estimation unit 104 may not be provided, and in that case, the loss completion unit may determine the correspondence relationship between the number of times the NC program is executed and an estimated waveform.

[0062] (Memory Unit 108) As already described, the memory unit 108 stores actual waveforms acquired from industrial machines and NC programs that control the machining operations of the industrial machines. The memory unit 108 also stores the correspondence between the actual waveforms and NC programs output from the matching unit 101, the correspondence between the number of NC program executions and the actual waveforms output from the correspondence analysis unit 103, and the correspondence between the estimated waveforms and NC programs output from the waveform generation unit 107. In addition to the actual waveforms and NC programs, the memory unit 108 also stores machine data, CNC (Computerized Numerical Control) parameters, macro variables, workpiece origin offsets, tool offset values, etc. Furthermore, the memory unit 108 stores servo waveform data such as cutting distance, cutting coefficient, estimated cutting force, cutting tool temperature, and torque command, as well as digitized workpiece (workpiece) shape and contact length, and tool information. The shape and contact length of the workpiece (workpiece) and the tool information are characteristic values ​​that indicate at least one characteristic of the workpiece and the tool. The memory unit 108 also stores servo waveform data of the estimated cutting force. The memory unit 108 stores the actual waveform characteristic amounts and estimated waveform characteristic amounts output from the characteristic amount calculation unit 102, the correspondence between the estimated waveform and the NC program output from the state estimation unit 104, and execution parameters such as cutting coefficients output from the parameter optimization unit 106.

[0063] (Waveform correlation unit 109) The waveform correlation unit 109 determines the correspondence relationship between the actual waveform and the estimated waveform (seventh correspondence relationship) from the correspondence relationship between the estimated waveform and the NC program (fourth correspondence relationship) and the correspondence relationship between the actual waveform and the NC program (fifth correspondence relationship), and stores the correspondence relationship in the memory unit 108.

[0064] The above describes the configuration of the data loss compensation device 10. Next, the operation of the data loss compensation configuration unit of the data loss compensation device 10 will be described using a flowchart.

[0065] 15 is a flowchart showing the missing data repair process of the missing data repair unit, and FIG. 16 is a flowchart showing the process of step S14 in FIG.

[0066] In step S11, the feature calculation unit 102 calculates the real waveform feature and the estimated waveform feature over a certain interval using the real waveform and the estimated waveform, respectively, and the parameter optimization unit 106 optimizes the execution parameters so that the fluctuations in the real waveform feature and the fluctuations in the estimated waveform feature coincide with each other.

[0067] In step S12, the correspondence analysis unit 103 determines whether there are any defects in the actual waveform for each number of NC program executions (including whether it is a unit of NC program or part of the NC program) based on the operation information, count information, and actual waveform, and determines the correspondence between the number of NC program executions and the actual waveform.

[0068] In step S13, the data loss complementation unit 105 acquires information from the memory unit 108 indicating whether or not there are any losses in the actual waveform for the number of times the NC program is executed. If there are any losses in the actual waveform, the process proceeds to step S14; if there are no losses in the actual waveform, the data loss complementation process is terminated.

[0069] In step S14, the waveform generation unit 107 generates an estimated waveform by simulation from the NC program, and the missing data complement unit 105 complements the missing part of the actual waveform using the estimated waveform estimated by simulation, thereby completing the missing data complement process. The process of step S14 will be further described below with reference to FIG. 16. As shown in FIG. 16, step S14 consists of steps S141 to S146.

[0070] In step S141, the missing data recovery unit 105 determines whether or not there is operation information. If there is operation information, the process proceeds to step S142, and if there is no operation information, the process proceeds to step S143.

[0071] In step S142, the missing data compensating unit 105 determines whether or not there is an estimated waveform. If there is an estimated waveform, the process proceeds to step S144, and if there is no estimated waveform, the process proceeds to step S143.

[0072] In step S143, the waveform generating unit 107 performs a simulation to generate an estimated waveform, and the process proceeds to step S144.

[0073] In step S144, the missing data complement unit 105 determines whether there is a missing part in the actual waveform, and if there is, determines whether the missing part is a part of the NC program or a whole NC program. If there is a missing part in the actual waveform in the NC program, the missing data complement unit 105 proceeds to step S145, and if there is a missing part in the actual waveform in the NC program, the missing data complement unit 105 proceeds to step S146.

[0074] In step S145, the missing data complementing unit 105 complements the missing part of the actual waveform with the estimated waveform, and ends the missing data complementing process. In step S146, the missing data complementing unit 105 complements the missing NC program unit of the actual waveform with the estimated waveform, and ends the missing data complementing process.

[0075] As described above, in order to realize the functional blocks included in the data loss repair device in this embodiment, the data loss repair device can be realized by hardware, software, or a combination of these. Here, being realized by software means being realized by a computer reading and executing a program.

[0076] In this embodiment, the functional blocks included in the data loss repair device are implemented by software or a combination thereof. Specifically, each data loss repair device includes a central processing unit (CPU) or other such processing device. The data loss repair device also includes a secondary storage device such as a hard disk drive (HDD) that stores various control programs such as application software or an operating system (OS), and a primary storage device such as a random access memory (RAM) that stores data temporarily required for the processing device to execute the programs.

[0077] In the data loss repair device, the arithmetic processing unit reads application software or an OS from the auxiliary storage device, and while loading the loaded application software or OS into the main storage device, performs arithmetic processing based on the application software or OS. Furthermore, based on the results of this calculation, various hardware components of each device are controlled. This realizes the functional blocks of this embodiment.

[0078] Each component included in the data loss repair device can be realized by hardware including electronic circuits, etc. When the data loss repair device is configured by hardware, some or all of the functions of each component included in the data loss repair device can be configured by an integrated circuit (IC), such as an ASIC (Application Specific Integrated Circuit), a gate array, an FPGA (Field Programmable Gate Array), or a CPLD (Complex Programmable Logic Device).

[0079] The program can be stored and supplied to a computer using various types of non-transitory computer-readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (e.g., hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, and semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (Random Access Memory)). The program may also be supplied to the computer by various types of transient computer readable media.

[0080] The data loss compensation device and data loss compensation method for industrial machinery according to the present embodiment described above can achieve the following effects: (1) It is possible to perform calculations, such as integration calculations, that assume that all data has been acquired. (2) It is only necessary to acquire data from necessary parts of the actual waveform where the losses have been compensated, minimizing the number of data acquisition points and reducing the network load. (3) Even if the amount of training data actually acquired is small due to losses, the training data can be increased by compensating for the losses.

[0081] Furthermore, according to the data loss repair device and data loss repair method for industrial machinery disclosed herein, including this embodiment, repair is possible by acquiring a certain amount of data once, without having to prepare multiple data sets that fill in the loss, as is the case with multiplexed repair.

[0082] The above-described embodiment is a preferred embodiment of the present invention, but the scope of the present invention is not limited to the above-described embodiment alone, and the present invention can be implemented in various modified forms within the scope that does not deviate from the gist of the present invention.

[0083] The following additional notes are further disclosed regarding the above embodiment. (Supplementary Note 1) A feature calculation unit (102) that calculates a first feature including a statistical amount of torque included in first servo waveform data based on actually measured first servo waveform data, and calculates a second feature including a statistical amount of torque included in the second servo waveform data based on second servo waveform data; a correspondence relationship analysis unit (103) that determines, based on the number of executions of an NC program that drives an industrial machine and the first servo waveform data, whether the first servo waveform data has any missing data in the number of executions and a first correspondence relationship between the number of executions and the first servo waveform data; a parameter optimization unit (106) that optimizes execution parameters including at least a cutting coefficient so that the first feature amount and the second feature amount coincide with each other, and determines a second correspondence relationship between the execution parameters and the first servo waveform data; and a waveform generation unit (107) that generates the second servo waveform data by a simulation using the NC program and the execution parameters, and determines a third correspondence relationship between the second servo waveform data and the execution parameters. a data defect complementing unit (105) that, when there is a defect in the first servo waveform data, complements the defect using the second servo waveform data.

[0084] (Supplementary Note 2) The data loss compensation device according to Supplementary Note 1, wherein the first servo waveform data and the second servo waveform data each include at least a feed axis position command, a spindle torque, and a spindle speed.

[0085] (Supplementary Note 3) The waveform generating unit (107) includes: a matching unit (101) that outputs a fourth correspondence relationship between the second servo waveform data and the NC program; a matching unit (101) that matches the first servo waveform data with the NC program and determines a fifth correspondence relationship between the first servo waveform data and the NC program; a state estimating unit (104) that determines a sixth correspondence relationship between the number of executions and the second servo waveform data from the first correspondence relationship, the second correspondence relationship, and the third correspondence relationship; a waveform correlating unit (109) that determines a seventh correspondence relationship between the first servo waveform data and the second servo waveform data from the fourth correspondence relationship and the fifth correspondence relationship; and a storage unit (108) that stores characteristic values ​​that indicate at least one characteristic of a workpiece and a tool, The data loss compensation device described in Appendix 1 or 2, wherein the waveform generation unit (107) is provided with a cutting force estimation unit (1074) that estimates a cutting force using a torque command included in the second servo waveform data and a cutting coefficient obtained based on the characteristic value.

[0086] (Appendix 4) The data defect completion unit (105) is a data defect completion device described in Appendix 3, wherein, when the correspondence relationship analysis unit (103) determines that the first servo waveform data has a first defect in the entire NC program, the data defect completion unit (105) completes the first defect using the second servo waveform data corresponding to the first defect from the fourth correspondence relationship, and when the correspondence relationship analysis unit (103) determines that the first servo waveform data has a second partial defect corresponding to a part of the NC program, the data defect completion unit (105) completes the second partial defect using the second servo waveform data corresponding to the second partial defect from the fourth correspondence relationship and the seventh correspondence relationship.

[0087] (Supplementary Note 5) The data loss complementation device according to Supplementary Note 1 or 2, wherein the waveform generation unit (107) comprises: an NC program interpretation unit (1071) that interprets an NC program that defines the operation of positioning the feed axis and speed control of the main spindle; a command generation unit (1072) that interpolates command points from the NC program interpreted by the NC program interpretation unit and generates a position command value or a speed command value; and a feedback control unit (1073) that performs feedback control to make the drive of the electric motor that drives the feed axis or the main spindle follow the position command value or the speed command value generated by the command generation unit.

[0088] (Supplementary Note 6) A computer performs the following processing: based on actually measured first servo waveform data, calculates a first feature amount including a statistic of torque included in the first servo waveform data; based on second servo waveform data, calculates a second feature amount including a statistic of torque included in the second servo waveform data; based on the number of times an NC program that drives an industrial machine is executed and the first servo waveform data, determines whether or not there is a defect in the first servo waveform data for the number of times the program is executed and a first correspondence relationship between the number of times the program is executed and the first servo waveform data; optimizes execution parameters including at least a cutting coefficient so that the first feature amount and the second feature amount match, and determines a second correspondence relationship between the execution parameters and the first servo waveform data; generates the second servo waveform data by a simulation using the NC program and the execution parameters, and determines a third correspondence relationship between the second servo waveform data and the execution parameters; and if there is a defect in the first servo waveform data, complements the defect using the second servo waveform data. A method for completing missing data in industrial machinery.

[0089] REFERENCE SIGNS LIST 10 Data missing complement device 101 Matching unit 102 Feature amount calculation unit 103 Correspondence relationship analysis unit 104 State estimation unit 105 Data missing complement unit 106 Parameter optimization unit 107 Waveform generation unit 108 Storage unit 1071 NC program interpretation unit 1072 Command generation unit 1073 Feedback control unit 1074 Cutting force estimation unit

Claims

1. A feature calculation unit that calculates a first feature including a statistical amount of torque included in first servo waveform data based on actually measured first servo waveform data, and calculates a second feature including a statistical amount of torque included in the second servo waveform data based on second servo waveform data; a correspondence relationship analysis unit that determines whether or not the first servo waveform data has a loss in the number of executions and a first correspondence relationship between the number of executions and the first servo waveform data based on the number of executions of an NC program that drives an industrial machine and the first servo waveform data; a parameter optimization unit that optimizes execution parameters including at least a cutting coefficient so that the first feature amount and the second feature amount match, and determines a second correspondence relationship between the execution parameters and the first servo waveform data; a waveform generation unit that generates the second servo waveform data by a simulation using the NC program and the execution parameters, and determines a third correspondence relationship between the second servo waveform data and the execution parameters; a data loss compensation unit that, when there is a loss in the first servo waveform data, compensates for the loss by using the second servo waveform data.

2. The data loss repair device according to claim 1, wherein the first servo waveform data and the second servo waveform data each include at least a feed axis position command, a spindle torque, and a spindle speed.

3. The data loss completion device as described in claim 1 or 2, wherein the waveform generation unit includes: a matching unit that outputs a fourth correspondence between the second servo waveform data and the NC program; matching the first servo waveform data with the NC program to determine a fifth correspondence between the first servo waveform data and the NC program; a state estimation unit that determines a sixth correspondence between the number of executions and the second servo waveform data from the first correspondence, the second correspondence, and the third correspondence; a waveform correlation unit that determines a seventh correspondence between the first servo waveform data and the second servo waveform data from the fourth correspondence and the fifth correspondence; and a memory unit that stores characteristic values ​​indicating at least one characteristic of a workpiece and a tool, wherein the waveform generation unit includes a cutting force estimation unit that estimates a cutting force using a torque command included in the second servo waveform data and the cutting coefficient determined based on the characteristic value.

4. The data loss repair device as described in claim 3, wherein, when the correspondence relationship analysis unit determines that the first servo waveform data has a first loss in the entire NC program, the data loss repair unit repairs the first loss using the second servo waveform data corresponding to the first loss based on the fourth correspondence relationship, and when the correspondence relationship analysis unit determines that the first servo waveform data has a second partial loss corresponding to a part of the NC program, the data loss repair unit repairs the second partial loss using the second servo waveform data corresponding to the second partial loss based on the fourth correspondence relationship and the seventh correspondence relationship.

5. The data loss repair device according to claim 1 or 2, wherein the waveform generation unit comprises: an NC program interpretation unit that interprets an NC program that specifies the operation of positioning the feed axis and speed control of the spindle; a command generation unit that performs interpolation of command points from the NC program interpreted by the NC program interpretation unit and generates a position command value or a speed command value; and a feedback control unit that performs feedback control to cause the drive of the electric motor that drives the feed axis or the spindle to follow the position command value or the speed command value generated by the command generation unit.

6. A computer performs a process of: calculating a first feature amount including a statistical amount of torque included in first servo waveform data based on actually measured first servo waveform data, and calculating a second feature amount including a statistical amount of torque included in the second servo waveform data based on second servo waveform data; determining whether or not there is a loss in the first servo waveform data in the number of executions and a first correspondence relationship between the number of executions and the first servo waveform data based on the number of executions of an NC program that drives an industrial machine and the first servo waveform data; optimizing execution parameters including at least a cutting coefficient so that the first feature amount and the second feature amount match, and determining a second correspondence relationship between the execution parameters and the first servo waveform data; generating the second servo waveform data by a simulation using the NC program and the execution parameters, and determining a third correspondence relationship between the second servo waveform data and the execution parameters; and if there is a loss in the first servo waveform data, complementing the loss using the second servo waveform data. The present invention relates to a method for completing missing data in industrial machinery.

Citation Information

Patent Citations

  • Numerical controller and machine tool

    JP2020181424A

  • Biological data processing apparatus, biological data processing method, and program

    JP2021047548A

  • Analyzer

    JP2021076998A

  • Numerical control device, machining system, numerical control method, and machining method

    JP7224541B1