Parameter adjustment device and parameter adjustment method

Abstract

The parameter adjustment device has a characteristic amount calculation section, an evaluation index calculation section, a first optimal solution search section, and a display control section. The characteristic amount calculation section calculates a characteristic amount of machining based on a simulation of the operation of a machine tool, which is a control target, according to a tool movement command. The evaluation index calculation section calculates one or more evaluation index values that evaluate a machining result based on the characteristic amount of machining. The first optimal solution search section uses a first learning result that estimates the evaluation index values based on a command value generation parameter set, estimates the evaluation index values corresponding to the command value generation parameter set for the first search, and searches for a plurality of command value generation parameter sets, i.e., command value generation parameter set candidates, that simultaneously optimize the respective evaluation index values using the estimated results. The display control section sets the command value generation parameter set candidates in a command value generation device that generates the tool movement command, and displays the characteristic amount of machining calculated when the command value generation device operates and the respective evaluation index values in association with each other on a display section.

Description

Technical Field

[0001] This invention relates to a parameter adjustment device and a parameter adjustment method for adjusting parameters related to command value generation in a command value generation apparatus that generates tool movement commands for driving a drive device of a machine tool based on a machining program. Background Technology

[0002] When machining a workpiece into a desired shape using a machine tool, a machining program is typically created using CAM (Computer-Aided Manufacturing). This program contains information related to the machined shape, the tool feed rate, and the tool spindle speed. Additionally, a command value generation device reads this program and calculates the toolpath by performing coordinate transformations, tool length correction, tool diameter correction, and mechanical error correction. The command value generation device also performs acceleration and deceleration processing, calculating the command points (interpolation points) on the toolpath per unit time. In most cases, a Numerical Control (NC) device is used in the command value generation device.

[0003] The instruction value generation device is equipped with various functions to enable faster and more precise machining operations performed by the machine tool. The operator needs to adjust numerous parameters related to these functions, depending on the shape and intended use of the workpiece, considering factors such as cycle time (machining time), surface shape accuracy (machining accuracy), and surface quality. Therefore, parameter adjustment of the instruction value generation device can be time-consuming or cumbersome, and may fail to achieve adjustments that match the operator's preferences. In this context, a technique for assisting parameter adjustment is disclosed in Patent Document 1, which involves executing a test program with multiple parameter settings to select the optimal set of parameters for evaluation indicators determined by machining accuracy and machining time.

[0004] Patent Document 1: Japanese Patent No. 5956619 Summary of the Invention

[0005] However, in the technology described in Patent Document 1, when the test procedure used for parameter adjustment differs from the shape of the workpiece processed by the operator, the accuracy of parameter adjustment deteriorates, and it becomes impossible to obtain an adjustment result that matches the operator's preferences. On the other hand, if it is necessary to maintain the accuracy of parameter adjustment in order to obtain an adjustment result that matches the operator's preferences, the test procedure must be changed or the adjustment range of each parameter must be corrected. This series of operations, starting from the parameter adjustment operation, needs to be repeated multiple times until it converges to the operator's preferences. In the past, when parameters converge to the operator's preferences, parameter adjustment operations needed to be performed through trial and error, thus imposing additional workload and time on the operator. Therefore, in parameter adjustment, a technology that can converge to the operator's preferences compared to the past is required.

[0006] The present invention was made in view of the above circumstances, and its object is to provide a parameter adjustment device that can make the parameters of the instruction values ​​that match the operator's preferences converge more quickly than before.

[0007] To address the aforementioned issues and achieve the objective, the parameter adjustment device of the present invention adjusts multiple parameters used in generating tool movement commands, namely, the command value generation parameter set. This tool movement command is composed of a set of interpolation points per unit time on the tool path calculated based on a machining program used to machine the workpiece. It includes a feature quantity calculation unit, an evaluation index calculation unit, a first optimal solution exploration unit, and a display control unit. The feature quantity calculation unit simulates the movement of the machine tool controlling the workpiece according to the tool movement command and calculates the machining feature quantities. The evaluation index calculation unit calculates an evaluation index value greater than or equal to one based on the machining feature quantities. The first optimal solution exploration unit uses a first learning result obtained by learning from the command value generation parameter set and evaluation index values ​​to infer evaluation index values. It infers evaluation index values ​​corresponding to the command value generation parameter set used in the first exploration and uses the inferred results to explore multiple command value generation parameter sets, namely, candidate command value generation parameter sets, that simultaneously optimize each evaluation index value. The display control unit sets the command value generation parameter set candidates to the command value generation device that generates tool movement commands, and displays the machining feature quantities and various evaluation index values ​​calculated when the command value generation device is activated on the display unit in association.

[0008] The effects of the invention

[0009] The parameter adjustment device involved in this invention has the effect that it can make the parameters for the instruction values ​​that match the operator's preferences converge more quickly than before. Attached Figure Description

[0010] Figure 1 This is a diagram illustrating an example of the structure of the parameter adjustment device involved in Embodiment 1.

[0011] Figure 2 This is a diagram showing an example of the shape of the processing target.

[0012] Figure 3 This is a diagram showing an example of the shape of the processing target.

[0013] Figure 4 This is a diagram showing an example of the shape of the processing target.

[0014] Figure 5 It means to Figures 2 to 4 The diagram shows an example of a machining procedure for machining the target shape.

[0015] Figure 6 This is a diagram illustrating an example of the change in acceleration / deceleration waveforms when allowable acceleration variations.

[0016] Figure 7 This is a diagram illustrating an example of how the movement path changes when the allowable path error varies.

[0017] Figure 8 This is a diagram illustrating an example of how acceleration / deceleration waveforms change when the allowable path error varies.

[0018] Figure 9 This is a diagram illustrating an example of how the tool's movement path changes when the filter time constant varies.

[0019] Figure 10 This is a diagram illustrating an example of how acceleration / deceleration waveforms change when the filter time constant varies.

[0020] Figure 11 This is a diagram illustrating the relationship between the machining error amount of the machining surface in the machining target shape and the machining time when machining actions are performed based on the parameter set generated from the instruction values ​​of groups 1 to 4.

[0021] Figure 12 This is a diagram illustrating an example of the neural network used in the learning process of Implementation Method 1.

[0022] Figure 13 This is a diagram showing an example of a set of command value generation parameters related to the machining surface explored by the first optimal solution exploration unit in Implementation 1.

[0023] Figure 14 It means about from Figure 13This diagram illustrates an example of setting preference information by the operator to generate a parameter set candidate by selecting one instruction value from the categories of machining time priority mode, machining accuracy priority mode, surface quality priority mode, and balance mode related to the machining surface.

[0024] Figure 15 This is a flowchart illustrating an example of the sequence of parameter adjustment methods involved in Implementation Method 1.

[0025] Figure 16 This is a diagram illustrating an example of the machining process for a component with a blade shape.

[0026] Figure 17 This is a diagram illustrating an example of the structure of the parameter adjustment device involved in Embodiment 2.

[0027] Figure 18 This is a flowchart illustrating an example of the sequence of parameter adjustment methods involved in Implementation Method 2.

[0028] Figure 19 This is a diagram illustrating an example of the structure of a computer system that implements the parameter adjustment device according to embodiments 1 and 2. Detailed Implementation

[0029] The parameter adjustment device and parameter adjustment method according to the embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] Implementation method 1.

[0031] Figure 1 This diagram illustrates an example of the structure of the parameter adjustment device according to Embodiment 1. The parameter adjustment device 1 is a device for adjusting a set of multiple parameters, or command value generation parameters, used to generate tool movement commands. These tool movement commands are composed of a set of interpolation points per unit time along the toolpath calculated based on a machining program used to machine the workpiece. The tool movement commands are commands used to drive a drive device such as a servo motor of a machine tool.

[0032] The instruction value generation device 3 outputs the tool movement instructions per unit time to the parameter adjustment device 1 according to the externally input machining program 310. The machining program 310 is a computer program that describes the tool path movement instructions corresponding to the machining target shape 320 and the current movement speed instructions. The tool path movement instructions specify the coordinate values ​​and the current movement mode through G codes such as G0 and G1, and the tool path movement speed instructions specify the speed values ​​through F codes.

[0033] The machining target shape 320 is the shape data of the target object to be machined, including the surface to be machined. The machining target shape 320 is externally input to the parameter adjustment device 1. In one example, the machining target shape 320 is input to the parameter adjustment device 1 through methods such as data transformation from CAD (Computer Aided Design) data, graphical input generated by operator operation via keyboard, etc.

[0034] Figures 2 to 4 This is a diagram showing an example of the shape of the processing target. Figure 2 This is a perspective view of the target shape 320 being processed. Figure 3 This is a front view of the target shape 320 being processed. Figure 4 This is a top view of the machining target shape 320. The machining target shape 320 has a hemispherical protrusion 322 on the upper surface 321a of the cuboid block 321, and has the shape after one corner of the upper surface 321a is removed by a plane. If focusing on the upper surface 321a, the machining target shape 320 has a machining surface S1 forming the hemispherical protrusion 322, a planar machining surface S2 forming the area of ​​the upper surface 321a other than the machining surface S1, and a machining surface S3 on the plane at the location where the corner is removed. In addition, there is an annular machining edge E1 at the boundary between the machining surfaces S1 and S2, and a straight machining edge E2 at the boundary between the machining surfaces S2 and S3.

[0035] Figure 5 It means to Figures 2 to 4 The diagram shows an example of a machining process for machining the target shape shown. In machining process 310, it is described that the upper surface 321a of the cuboid block 321 is made into... Figures 2 to 4 The method shown is for processing the machine tool's motion in the way the target shape 320 is shaped. Figure 5 The example shown illustrates the processing path for scanline machining, but contour machining could also be used. Furthermore, the machining direction is not limited.

[0036] When the instruction value generation device 3 outputs the tool movement instructions per unit time according to the externally input machining program 310, it performs analysis processing, acceleration / deceleration processing, smoothing processing, and interpolation processing. Analysis processing outputs the movement path and feed rate along the movement path based on the machining program 310. Acceleration / deceleration processing calculates the acceleration / deceleration waveform between the stop state and the feed rate state based on a preset allowable acceleration. Smoothing processing outputs the movement instructions after smoothing the movement path based on a preset allowable path error and acceleration / deceleration waveform. Smoothing processing smooths the speed waveform after smoothing. Smoothing processing is also known as moving average filtering. Interpolation processing calculates the tool position per unit time, i.e., the interpolation point, when moving at the smoothed speed. Here, each tool movement instruction per unit time is called an interpolation point.

[0037] Each process in the instruction value generation device 3 operates according to parameters. The parameters will be explained below.

[0038] In acceleration / deceleration processing, the acceleration / deceleration waveform changes with a set allowable acceleration. That is, the allowable acceleration becomes a parameter. Figure 6 This is a graph illustrating an example of the changes in acceleration / deceleration waveforms when the allowable acceleration varies. In this graph, the horizontal axis represents time, and the vertical axis represents velocity. The graph represented by solid lines shows the acceleration / deceleration waveform when the allowable acceleration is high, and the graph represented by dashed lines shows the acceleration / deceleration waveform when the allowable acceleration is low. According to... Figure 6 It can be seen that by reducing the allowable acceleration, a smoother acceleration / deceleration waveform with lower acceleration can be obtained compared with that with a high allowable acceleration, but the processing time increases.

[0039] In the smoothing process, the movement command and speed waveform change according to the set allowable path error. That is, the allowable path error becomes a parameter. Figure 7 This is a diagram illustrating an example of how the movement path changes when the allowable path error varies. Figure 8 This is a graph illustrating an example of how acceleration / deceleration waveforms change when the allowable path error varies. Figure 8 In the diagram, the horizontal axis represents time, and the vertical axis represents velocity. Figure 7 In machining program 310, the tool's movement path, as shown by the dashed line, travels along the X-axis and then along the Y-axis. The movement path represented by the solid line is the path with a large allowable path error, while the movement path represented by the dashed line is the path with a small allowable path error. Figure 8 In the middle, it is shown that along Figure 7The text describes the acceleration and deceleration waveforms along the X-axis and Y-axis during machining along the movement path. Specifically, regarding the Y-axis acceleration and deceleration waveforms, solid lines indicate cases with large permissible path errors, while dashed lines indicate cases with small permissible path errors. Based on... Figure 7 and Figure 8 It can be seen that as the allowable path error increases, the machining time can be shortened compared to when the allowable path error is small, but the tool path error increases.

[0040] In the smoothing process, the tool movement command and speed waveform change in a smooth manner by setting the time constant of the moving average filter. Hereinafter, the time constant of the moving average filter is referred to as the filtering time constant. In this process, the filtering time constant becomes a parameter. The interpolation point x on the path after the moving average filter, that is, the path of the tool movement command, is represented by the average value of the points X on the path before the moving average filter, that is, the path of the machining program 310, and can therefore be represented by the following equation (1).

[0041] Formula 1

[0042]

[0043] Here, n represents the interpolation point number from the starting point to the ending point. Additionally, m is the filtering time constant of the moving average filter, which is set via parameters.

[0044] Figure 9 This is a diagram illustrating an example of how the tool's movement path changes with variations in the filter time constant. Figure 10 This is a graph illustrating an example of how acceleration / deceleration waveforms change with variations in the filter time constant. Figure 10 In the diagram, the horizontal axis represents time, and the vertical axis represents velocity. Figure 9 In machining program 310, the tool's movement path, as shown by the dashed lines, travels along the X-axis and then along the Y-axis. The movement path represented by solid lines is the path when the filter time constant is small, and the movement path represented by dashed lines is the path when the filter time constant is large. Figure 10 In the middle, it is shown that along Figure 9 The text describes the acceleration / deceleration waveforms along the X-axis and Y-axis during machining along the movement path. Additionally, dashed lines indicate cases with a large filter time constant, while solid lines indicate cases with a small filter time constant. Based on... Figure 9 and Figure 10 It can be seen that by increasing the filtering time constant of the moving average filter, a smooth acceleration / deceleration waveform can be obtained by comparing it with a smaller filtering time constant, but the machining time and tool path error become larger.

[0045] As described below, in Embodiment 1, the parameter adjustment device 1 processes a total of three parameters—allowable acceleration, allowable path error, and filter time constant—to generate a parameter set as instruction values. That is, the parameter adjustment device 1 uses this three instruction values ​​to generate a parameter set as the object of parameter adjustment. However, it is not limited to the three parameters processed in Embodiment 1; it is also possible to process all parameters that affect the interpolation point generated from the instruction value generation device 3 as the object of parameter adjustment.

[0046] Furthermore, the instruction value generation device 3 operates based on the setting values ​​of the instruction value generation parameter set pre-stored in the setting value storage unit of the instruction value generation device 3. The setting values ​​in the setting value storage unit can be rewritten via external input from the parameter adjustment device 1.

[0047] return Figure 1 The parameter adjustment device 1 includes a feature quantity calculation unit 11, an evaluation index calculation unit 12, an evaluation index information storage unit 13, a first optimal solution exploration unit 14, a candidate information storage unit 15, a preference information setting unit 16, a display unit 17, a second optimal solution exploration unit 18, and an adjusted instruction value generation parameter set storage unit 19.

[0048] The feature quantity calculation unit 11 simulates the movement of the machine tool controlled by the control object based on the tool movement command generated by the command value generation device 3, and calculates the machining feature quantities. An example of the machining feature quantities is the distance between the machining target shape 320 and the tool positioned at the tool tip point, i.e., the machining error; the speed of the tool tip point; the acceleration of the tool tip point; the jerk of the tool tip point; the position of each of the multiple drive axes of the machine tool; the speed of each of the multiple drive axes of the machine tool; the acceleration of each of the multiple drive axes of the machine tool; the jerk of each of the multiple drive axes of the machine tool; and the reverse position of each of the multiple drive axes of the machine tool.

[0049] Here, when processing the entire workpiece under the same single condition, the process is as described above. However, it is also possible to divide the workpiece into multiple parts and process each part by changing the conditions. In this case, the feature quantity calculation unit 11 calculates the tool tip point by simulating the operation of the machine tool controlled by the command value generation device 3 based on the tool movement command generated by the command value generation device 3. For each machining surface or machining edge of the target shape 320 that has one or more machining surfaces or machining edges, the processing information at the tool tip point, i.e., the processing feature quantity, is calculated. The target shape 320 has one or more machining surfaces or machining edges that correspond to shape structure elements.

[0050] The feature quantity calculation unit 11 first performs a tool tip point estimation process to estimate the tool tip point, and then performs a feature quantity calculation process to calculate the machining features at the tool tip point. The tool tip point estimation process and the feature quantity calculation process will be explained below.

[0051] <Tool tip point estimation processing>

[0052] In order to follow the tool movement command generated by the command value generation device 3, the feature quantity calculation unit 11 estimates the tool tip point using the result information obtained from the drive control unit of the machine tool, which is driven by the actual control object or by simulation. When using the simulation results, the feature quantity calculation unit 11 simulates the movement of the machine tool on a computer and estimates the actual tool tip point based on the output of the command value generation device 3, i.e., the interpolation point. Specifically, the machine tool's movement is simulated by pre-setting parameters such as inertia, viscosity, and elasticity of the machine tool; the resonant or anti-resonant frequency caused by inertia, viscosity, and elasticity; parameters of backlash or idle rotation during shaft reversal; parameters of thermal displacement; and parameters of displacement caused by reaction force during machining. Here, the estimation accuracy of the tool tip point calculated by simulation can be varied. In one example, if an estimation accuracy equivalent to that of the machine tool's drive shaft is obtained, the drive shaft position information can be used as the tool tip point; if an estimation accuracy equivalent to that of the interpolation point is obtained, the interpolation point can be used as the tool tip point. Furthermore, by utilizing the results of actual actions, the feature quantity calculation unit 11 operates the actual working machine to obtain information equivalent to the tip point of the tool.

[0053] <Characteristic Quantity Calculation and Processing>

[0054] The feature quantity calculation unit 11 calculates the machining feature quantity at the tool tip point obtained through the tool tip point estimation process, and associates it with the machining surface or machining edge of the machining target shape 320. The following describes the calculation methods for machining error, tool tip point speed, tool tip point acceleration, tool tip point jerk, the positions of the multiple drive shafts of the machine tool, the speeds of the multiple drive shafts of the machine tool, the accelerations of the multiple drive shafts of the machine tool, the jerk of the multiple drive shafts of the machine tool, and the reverse positions of the multiple drive shafts of the machine tool, as examples of machining feature quantities.

[0055] The machining error can be calculated as the shortest distance between the position of the cutting point corresponding to the tool tip and the shape surface of the tool configured according to the position of the tool tip and the tool direction. The position of the tool tip is calculated based on information obtained from simulating the behavior of the working machine as the controlled object, or by operating the controlled object.

[0056] The velocity, acceleration, and jerk of the tool tip can be calculated as follows. For the tool tip points from the starting point to the ending point, if the position of the nth tool tip point is set as PT(n), and the position of the (n+1)th tool tip point advancing at a specified control cycle time Δt is set as PT(n+1), then the velocity VT(n) of the nth tool tip point is calculated as shown in equation (2) by dividing the distance between the positions PT(n+1) and PT(n) of the two tool tip points by the specified control cycle time Δt.

[0057] Formula 2

[0058]

[0059] Similarly, the acceleration AT(n) of the nth tool tip point is calculated by dividing the difference between the velocities VT(n+1) and VT(n) at the two tool tip points by the time Δt of the specified control cycle, as shown in Equation (3).

[0060]

Formula 3

[0061]

[0062] Similarly, the acceleration JT(n) of the nth tool tip point is calculated by dividing the difference between the accelerations AT(n+1) and AT(n) of the two tool tip points by the time Δt of the specified control cycle, as shown in Equation (4).

[0063]

Formula 4

[0064]

[0065] Furthermore, the positions, speeds, accelerations, and jerks of the various drive axes of the machine tool can be calculated as follows. The position PM1(n) of the first drive axis corresponding to the nth tool tip point can be obtained from the timing data of the machine tool's operation information. Operation information represents the operating state of the machine tool. The operation information includes information obtained from the machine tool, the CNC device controlling the machine tool (i.e., the command value generation device 3), or sensors installed on the machine tool. In this example, the operation information includes the position data of the various drive axes of the machine tool.

[0066] If the position of the first drive axis corresponding to the (n+1)th tool tip point that advances at a specified control cycle time Δt is set as PM1(n+1), then the speed of the first drive axis corresponding to the nth tool tip point at time t is calculated by the following formula (5).

[0067]

Formula 5

[0068]

[0069] Similarly, if the speed of the first drive axis corresponding to the (n+1)th tool tip point that advances at a specified control cycle time Δt is set as VM1(n+1), then the acceleration AM1(n) of the first drive axis corresponding to the nth tool tip point is calculated by the following equation (6).

[0070]

Formula 6

[0071]

[0072] Similarly, if the acceleration of the first drive axis corresponding to the (n+1)th tool tip point that advances with a specified control cycle time Δt is set as AM1(n+1), then the acceleration JM1(n) of the first drive axis corresponding to the nth tool tip point is calculated by the following formula (7).

[0073]

Formula 7

[0074]

[0075] Furthermore, for other drive shafts besides the first drive shaft, the position, velocity, acceleration, and jerk can also be calculated using the same method.

[0076] The reverse positions of the multiple drive axes of the machine tool can be calculated as follows. Using the method described above, the speed VM1(n) of the first drive axis corresponding to the nth tool tip point and the speed VM1(n+1) of the first drive axis corresponding to the (n+1)th tool tip point advancing over a specified control cycle time Δt are calculated. Then, by comparing the signs of speeds VM1(n) and VM1(n+1), the position corresponding to the moment the signs reverse can be set as the reverse position of the first drive axis. Furthermore, the reverse positions of the other drive axes besides the first drive axis can also be determined using the same method.

[0077] The feature calculation unit 11 associates the processed feature quantities with the processed surface or processed edge of the processed target shape 320. In one example, for each piece of information of the processed surface and processed edge in the processed target shape 320, identification information, i.e., ID number, is pre-assigned to identify the processed surface and processed edge, thereby enabling the determination of the processed feature quantity for the corresponding ID number.

[0078] The feature quantity calculation unit 11 outputs the machining feature quantities calculated in the above manner to the evaluation index calculation unit 12. When machining the entire workpiece under one condition, the feature quantity calculation unit 11 outputs the machining feature quantities related to the machining target shape 320 to the evaluation index calculation unit 12. In addition, when the entire workpiece is divided into multiple parts and each separate part is machined under different conditions, the feature quantity calculation unit 11 outputs the machining feature quantities separately for each machining surface or each machining edge of the machining target shape 320 to the evaluation index calculation unit 12.

[0079] The evaluation index calculation unit 12 calculates an evaluation index value greater than or equal to one based on the processing feature quantities calculated by the feature quantity calculation unit 11. In the following description, examples are given of processing time as cycle time, processing accuracy as shape accuracy of the processed surface, and surface quality as surface accuracy of the processed surface. Furthermore, processing time, processing accuracy, and surface quality are subject to trade-offs.

[0080] In one example, the evaluation index value Qt related to machining time can be calculated using the deceleration rate of the tool tip point, which is calculated based on the result information of the instruction speed described in machining program 310, by the following formula (8).

[0081]

Form 8

[0082]

[0083] Here, 〇 represents the machining surface or machining edge of the machining target shape 320. Figures 2 to 4 In the case of a machining target shape of 320, 〇 represents the machining surfaces S1-S3 and the machining edges E1 and E2. Additionally, N represents the number of tool tip points corresponding to each machining surface and machining edge, and F... c This indicates the command speed, and F indicates the speed of the tool tip.

[0084] According to equation (8), the velocity F of the tool tip is equal to the commanded velocity F. cThe more consistent the values, the smaller the evaluation index value Qt for processing time becomes. That is, in Implementation 1, it can be said that the smaller the evaluation index value Qt is, the better the command value generation parameter set in parameter adjustment device 1 is in terms of processing time. However, the evaluation index value Qt only needs to be able to evaluate the processing time, and is not limited to being determined by equation (8). In one example, the evaluation index value Qt can be the number of data points N corresponding to the tool tip points of a specific machining surface and machining edge, or it can be the time calculated by multiplying the number of data points N by the execution unit.

[0085] In addition, in equation (8), the speed of the tool tip point is used when calculating the evaluation index value Qt of the machining time, but the average speed of the tool tip point, the maximum speed, the average speed of the speed of each of the multiple drive axes of the machine tool, and the maximum speed can also be used.

[0086] Furthermore, generally speaking, when machining time decreases, there is a tendency for the average value of acceleration, the maximum value of acceleration, the average value of jerk, and the maximum value of jerk to increase. Therefore, the average value of acceleration at the tool tip, the maximum value of acceleration, the average value of jerk, the maximum value of jerk, the average value of acceleration of each of the multiple drive axes of the machine tool, the maximum value of acceleration, the average value of jerk, and the maximum value of jerk can be used as the evaluation index value Qt for machining time. However, in this case, the larger the evaluation index value Qt is, the better the parameter set for generating the command value in parameter adjustment device 1 is determined from the perspective of machining time.

[0087] In one example, the evaluation index value Qa related to machining accuracy can be calculated using the average value of the machining error amount between the machining target shape 320 and the tool positioned at the tool tip point, i.e., the distance between the tool and the tool tip point.

[0088]

Form 9

[0089]

[0090] Here, 〇 represents the machining surface or machining edge of the machining target shape 320. Figures 2 to 4 In the case of a machining target shape of 320, 〇 represents the machining surfaces S1-S3 and the machining edges E1 and E2. In addition, N represents the number of data points of the tool tip corresponding to each machining surface and machining edge, and e represents the machining error calculated as a machining feature quantity.

[0091] According to equation (9), the smaller the machining error e becomes, the smaller the evaluation index value Qa of the machining accuracy becomes. That is, in embodiment 1, it can be said that the smaller the evaluation index value Qa is, the better the command value generation parameter set in the parameter adjustment device 1 is in terms of machining accuracy. However, the evaluation index value Qa only needs to be able to evaluate the machining accuracy and is not limited to being determined by equation (9). In one example, it can be a value representing the degree of mechanical vibration or the tool's following ability.

[0092] In addition, in equation (9), when calculating the evaluation index value Qa of machining accuracy, the machining error amount corresponding to the machining surface and machining edge is used, but the maximum and minimum values ​​of the machining error amount corresponding to the specific machining surface and machining edge can also be used.

[0093] In addition, generally speaking, when machining accuracy deteriorates, there is a tendency for the average value of acceleration, the maximum value of acceleration, the average value of jerk, and the maximum value of jerk to increase. Therefore, the maximum and minimum values ​​of acceleration at the tool tip point, the maximum and minimum values ​​of jerk at the tool tip point, the maximum and minimum values ​​of acceleration of each of the multiple drive axes of the machine tool, and the maximum and minimum values ​​of jerk of each of the multiple drive axes of the machine tool can also be used as the evaluation index value Qa of machining accuracy.

[0094] Furthermore, the larger the evaluation index value Qa is, the better the parameter set generated by the instruction value in the parameter adjustment device 1 can be determined from the perspective of machining accuracy.

[0095] In one example, the evaluation index value Qq related to surface quality can be calculated using the variance of the machining error amount, which is the distance between the machining target shape 320 and the tool positioned at the tool tip point, by the following formula (10).

[0096]

Formula 10

[0097]

[0098] Here, 〇 represents the machining surface or machining edge of the machining target shape 320. Figures 2 to 4 In the case of a machining target shape of 320, 〇 represents the machining surfaces S1-S3 and the machining edges E1 and E2. Additionally, N represents the number of tool tip points corresponding to each machining surface and edge, and e represents the machining error calculated as a machining feature quantity. a This represents the average value of the machining error.

[0099] According to equation (10), the smaller the variance of the processing error e becomes, the smaller the evaluation index value Qq of the surface quality becomes. That is, in embodiment 1, it can be said that the smaller the evaluation index value Qq is, the better the parameter set for generating the command value in the parameter adjustment device 1 is in terms of surface quality. However, the evaluation index value Qq only needs to be able to evaluate the surface quality and is not limited to being determined by equation (10). In one example, it can be a value representing the degree of mechanical vibration.

[0100] In addition, in equation (10), when calculating the evaluation index value Qq of the surface quality, the difference between the machining error amount and the average value of the machining error amount corresponding to the machining surface and the machining edge is used. However, the difference between the maximum and minimum values ​​of the machining error amount corresponding to the specific machining surface and the machining edge, the difference between the maximum and minimum values ​​of the acceleration of the tool tip point, the difference between the maximum and minimum values ​​of the acceleration of the tool tip point, the difference between the maximum and minimum values ​​of the acceleration of the multiple drive shafts of the working machine, and the difference between the maximum and minimum values ​​of the acceleration of the multiple drive shafts of the working machine are also used.

[0101] Furthermore, the larger the evaluation index value Qq is, the better the parameter set generated by the instruction value in the parameter adjustment device 1 can be determined from the aspect of surface quality.

[0102] In addition, three evaluation indicators related to processing time, processing accuracy, and surface quality are calculated here, but any one or more of the three evaluation indicators can also be calculated to match the operator's preferences.

[0103] The evaluation index information storage unit 13 stores evaluation index information related to the processing surface or processing edge of the processing target shape 320, which is associated with the evaluation index values ​​and command values ​​generated by the evaluation index calculation unit 12 that are related to processing time, processing accuracy, and surface quality. Furthermore, the evaluation index information may also have corresponding processing characteristic quantities based on the evaluation index values ​​and command values ​​generated by the parameter set related to processing time, processing accuracy, and surface quality.

[0104] The first optimal solution exploration unit 14 uses the first learning result obtained from learning the instruction value generation parameter set based on the instruction value generation parameter set and evaluation index values ​​to infer the evaluation index values. It infers the evaluation index values ​​corresponding to the instruction value generation parameter set used in the first exploration, and uses the inferred results to explore multiple instruction value generation parameter set candidates that simultaneously optimize each evaluation index value. Furthermore, when exploring multiple instruction value generation parameter set candidates, the exploration focuses on the balance of evaluation index values ​​that are in a compromise relationship. In one example, the proportions of each evaluation index value relative to the sum of evaluation index values ​​for processing time, processing accuracy, and surface quality become balanced. Additionally, in one example, a situation where at least one of the three evaluation index values ​​deviating from the instruction value generation parameter set candidate is determined to be more than the proportion of the evaluation index value corresponding to other instruction value generation parameter set candidates is referred to as a difference in the balance of evaluation index values. In this example, the first optimal solution exploration unit 14 explores candidate parameter sets that simultaneously minimize one instruction value greater than or equal to each evaluation index value. Here, simultaneous minimization means finding, among the three evaluation index values ​​that are in a trade-off relationship, a solution where improving one evaluation index value would worsen the other objective functions.

[0105] The following section explains the learning process for learning the relationship between the parameter set for generating instruction values ​​and the evaluation index values, as well as the exploratory process for exploring the parameter set, using the learning results.

[0106] <Learning Processing>

[0107] In the learning process, the first optimal solution exploration unit 14 takes the evaluation index values ​​and parameter ranges related to processing time, processing accuracy, and surface quality as input, learns the relationship between the command value generation parameter set and the evaluation index values ​​calculated in the evaluation index calculation unit 12, and outputs the learning result. That is, the first optimal solution exploration unit 14 uses learning data including the command value generation parameter set and the evaluation index values ​​related to processing time, processing accuracy, and surface quality to generate a first learning result for inferring the evaluation index values ​​based on the command value generation parameter set.

[0108] Specifically, a neural network is constructed that takes the instruction value generation parameter set as input and the evaluation index value as output. The first optimal solution exploration unit 14 learns by updating the weighting coefficients of the neural network. Having learned by updating the weighting coefficients, the neural network outputs a well-predicted value of the evaluation index value corresponding to the instruction value generation parameter set. The first optimal solution exploration unit 14 uses the neural network to obtain a function that takes the instruction value generation parameter set as input and the evaluation index value as output, thereby obtaining the relationship between the instruction value generation parameter set and the evaluation index value, i.e., the first learning result.

[0109] The first optimal solution exploration unit 14 selects and outputs a set of command value generation parameters for performing the next machining operation from a specified parameter range within the machining target shape 320. When selecting the next set of command value generation parameters, the first optimal solution exploration unit 14 can select a set of command value generation parameters representing excellent evaluation index values ​​based on learning results, or it can select each set of command value generation parameters sequentially from points of an equally spaced grid. The first optimal solution exploration unit 14 has the function of updating the function for calculating evaluation index values ​​related to machining time, machining accuracy, and surface quality based on the command value generation parameter set.

[0110] Here, the process of performing the actions of the first optimal solution exploration unit 14 four times, up to the fourth set of command value generation parameter sets, is explained. The first set of command value generation parameter sets is labeled Pr1, the second set is labeled Pr2, the third set is labeled Pr3, and the fourth set is labeled Pr4. Each of the four sets of command value generation parameter sets has three parameters: allowable acceleration, allowable path error, and filtering time constant.

[0111] Figure 11 This is an example diagram showing the relationship between the machining error of the machined surface in the machined target shape and the machining time when performing machining actions based on the parameter set generated from the instruction values ​​of groups 1 to 4. Figure 11The diagram shows a mapping of machining errors for the machined surface S1. Mapping Ma shows the machining error and machining time for surface S1 when the parameter set is generated using the first set of command values. Mapping Mb shows the machining error and machining time for surface S1 when the parameter set is generated using the second set of command values. Mapping Mc shows the machining error and machining time for surface S1 when the parameter set is generated using the third set of command values. Mapping Md shows the machining error and machining time for surface S1 when the parameter set is generated using the fourth set of command values. Furthermore, the distribution of machining errors in surface S1 indicates surface quality. In surface S1, if the machining error is uniform, the surface quality is high; if the machining error is uneven, the surface quality is considered low. The shading added to the machined surface S1 in these mappings Ma-Md indicates the error amount, and the legend of the error amount is shown as "Error Amount" on the right side of each mapping Ma-Md. In addition, the processing time is represented by the slider in the "Production Cycle" section on the right side of each mapping map Ma-Md.

[0112] If the first optimal solution exploration unit 14 receives the evaluation index values ​​Qt1, Qa1, and Qq1 related to the machining surface S1 in the machining target shape 320 obtained by machining actions using the instruction value generation parameters as the first set of instruction value generation parameter set Pr1, then the first set of instruction value generation parameter set Pr1 is changed to the second set of instruction value generation parameter set Pr2. At this time, the second set of instruction value generation parameter set Pr2 can be selected based on the result of the machining action using the first set of instruction value generation parameter set Pr1, or it can be selected as preset, regardless of the result of the machining action using the first set of instruction value generation parameter set Pr1.

[0113] The first optimal solution exploration unit 14 receives the evaluation index values ​​Qt2-Qt4, Qa2-Qa4, and Qq2-Qq4 corresponding to the instruction value generation parameter sets Pr2-Pr4 from the second to the fourth group, in the same order as the first group of instruction value generation parameter set Pr1.

[0114] In obtaining Figure 11 Given the processing characteristics shown, from the perspective of processing time, among the four evaluation index values ​​Qt1-Qt4, the evaluation index value Qt1 is the smallest. That is, it can be said that the first set of instruction value generation parameters Pr1 is the instruction value generation parameter set that prioritizes processing time.

[0115] In terms of machining accuracy, among the four evaluation index values ​​Qa1-Qa4, the evaluation index value Qa2 is the smallest. That is, it can be said that the second set of command value generation parameter set Pr2 is a command value generation parameter set that prioritizes machining accuracy.

[0116] From the perspective of surface quality, among the four evaluation index values ​​Qq1-Qq4, the evaluation index value Qq3 is the smallest. That is, it can be said that the third set of instruction value generation parameter set Pr3 is the instruction value generation parameter set that prioritizes surface quality.

[0117] Furthermore, it can be said that the fourth set of instruction value generation parameters, Pr4, is a set of instruction value generation parameters that achieves a balance in all aspects of processing time, processing accuracy, and surface quality.

[0118] As described above, the first optimal solution exploration unit 14 repeatedly performs the action of obtaining the evaluation index value corresponding to the instruction value generation parameter. The first optimal solution exploration unit 14 uses the instruction value generation parameter and the evaluation index value corresponding to the instruction value generation parameter obtained by repeated implementation as learning data and performs a learning action using a neural network. Figure 12 This diagram illustrates an example of the neural network used in the learning process of Implementation Method 1. The neural network has an input layer, intermediate layers, and an output layer. Instruction values ​​are input to the input layer on the left to generate a parameter set, and evaluation index values ​​are output from the output layer on the right. The weighting coefficients for each node in the intermediate layer can be independently set from each node in the input layer, but... Figure 12 They are all labeled with the same weighting coefficient W1. Similarly, each node in the intermediate layer can independently set the weighting coefficients for each node in the output layer, but... Figure 12 All are marked with the same weighting coefficient W2.

[0119] The output values ​​of each node in the input layer are multiplied by a weighting coefficient W1, and the linear combination of these multiplications is input to the nodes in the intermediate layer. Similarly, the output values ​​of each node in the intermediate layer are multiplied by a weighting coefficient W2, and the linear combination of these multiplications is input to the nodes in the output layer. At each node in each layer, in one example, the output value can be calculated based on the input value using a non-linear function such as a sigmoid function. In both the input and output layers, the output value can be a linear combination of the input values.

[0120] The first optimal solution exploration unit 14 uses instruction values ​​to generate a parameter set and evaluation index values, and calculates the weighting coefficients W1 and W2 of the neural network. The weighting coefficients W1 and W2 of the neural network can be calculated using either backpropagation or gradient descent. In one example, the neural network learns by adjusting the weighting coefficients W1 and W2 in a way that the output from the output layer after inputting the instruction value parameter set to the input layer approximates the evaluation index value. However, the calculation method for the weighting coefficients W1 and W2 is not limited to the methods described above, even if the method used to obtain the weighting coefficients of the neural network is different.

[0121] If the weighting coefficients W1 and W2 of the neural network are determined, the relationship between the instruction value generation parameters and the evaluation index values ​​will be obtained. This concludes the example of learning using a 3-layer neural network. Learning using neural networks is not limited to the example above.

[0122] Through the above actions, the first learning result is obtained, which takes the parameter set generated by the neural network (i.e., the instruction value generation parameter set) as input and the evaluation index value as output. This first learning result is used to infer the evaluation index value based on the parameter generated from the instruction value.

[0123] If the first learning result is used, even without performing machining actions on the parameter set generated for the new instruction value, it is possible to obtain evaluation index values ​​Qt, Qa, and Qq for machining time, machining accuracy, and surface quality corresponding to the parameter set generated for the new instruction value.

[0124] Furthermore, in Implementation 1, a neural network was used to construct the relationship between the instruction value generation parameter set and the evaluation index value. However, if the relationship between the instruction value generation parameter set and the evaluation index value can be obtained, methods other than neural networks can also be used. In one example, to obtain the relationship between the instruction value generation parameter set and the evaluation index value, a simple function such as a quadratic polynomial or a probabilistic model such as a Gaussian process model can be used.

[0125] Furthermore, the prediction accuracy of the first learning result depends on the number of repetitions of the action to obtain the evaluation index value corresponding to the parameter set generated from the instruction value. With fewer repetitions, the first learning result can be obtained in a short time, but the error contained in the evaluation index value predicted based on the parameter set generated from the instruction value tends to increase. On the other hand, with a sufficient number of repetitions, the error contained in the evaluation index value predicted based on the parameter set generated from the instruction value decreases, but a longer time is required to obtain a first learning result with good accuracy.

[0126] When performing learning actions, if the evaluation index values ​​related to processing time, processing accuracy, and surface quality are near the boundary of the distribution of maximum or minimum evaluation index values, it is sufficient to ensure a sufficient number of repetitions for actions that obtain evaluation index values ​​corresponding to the instruction value generation parameter set. On the other hand, for areas outside the aforementioned boundary, it is not necessary to ensure a sufficient number of repetitions for actions that obtain evaluation index values ​​corresponding to the instruction value generation parameter set. In this case, as long as the evaluation index values ​​are obtained within a wide range of the specified parameters, a fewer number of repetitions is acceptable.

[0127] Furthermore, in this example, learning processing is performed for each machining surface or each machining edge in the machining target shape 320, but learning processing can also be performed simultaneously for multiple machining surfaces and machining edges. Figures 2 to 4 In the example of the machining target shape 320 shown, when machining surfaces S1 and S2 and the machining edge E1 enclosed by machining surfaces S1 and S2 are processed simultaneously, the result obtained by linearly combining the evaluation index values ​​of machining surfaces S1, S2 and machining edge E1 is used as the new evaluation formula Q'. If the evaluation index value of machining surface S1 is set to Q(S1), the evaluation index value of machining surface S2 is set to Q(S2), and the evaluation index value of machining edge E1 is set to Q(E1), then a S1 a S2 a E1 Let it be a coefficient, then the new evaluation formula Q' is expressed by the following formula (11).

[0128]

Formula 11

[0129] Q′ □ =a S1 Q □ (S1)+a S2 Q □ (S2)+a E1 Q □ (E1) ·…(11)

[0130] Here, □ represents the processing time, processing accuracy, or surface quality that is being evaluated. That is, the evaluation formula Q' represents the evaluation index values ​​of multiple processed surfaces and edges of the processed target shape 320. Figures 2 to 4 The example shows evaluation index values ​​for any of the following: machining time, machining accuracy, and surface quality, which are used as evaluation objects for machining surfaces S1-S3 and machining edges E1 and E2. Therefore, even when machining shape and structural elements composed of multiple machining surfaces or machining edges using a parameter set generated by a single instruction value, learning processing is still possible.

[0131] The learning data used by the first optimal solution exploration unit 14 during the learning process is data related to the control object used for calculating feature quantities by the feature quantity calculation unit 11.

[0132] <Exploratory Processing>

[0133] In the exploration process, the first optimal solution exploration unit 14 uses the first learning result for inferring evaluation index values ​​based on the command value generation parameter set to infer evaluation index values ​​corresponding to the command value generation parameter set used for exploration. Furthermore, the first optimal solution exploration unit 14 explores command value generation parameter set candidates that simultaneously optimize each evaluation index value using the inferred results. When the entire workpiece is processed under one condition, the first optimal solution exploration unit 14 explores command value generation parameter set candidates with respect to the processing target shape 320. On the other hand, when the entire workpiece is divided into multiple parts and each separate part is processed under different conditions, the first optimal solution exploration unit 14 explores command value generation parameter set candidates for each processing surface or each processing edge of the processing target shape 320. The command value generation parameter set candidates can be a single command value generation parameter set that simultaneously optimizes each evaluation index value, or multiple command value generation parameter sets. The command value generation parameter set used by the first optimal solution exploration unit 14 for exploration corresponds to the command value generation parameter set used in the first exploration.

[0134] That is, the first optimal solution exploration unit 14, based on the relationship between the command value generation parameters and the evaluation index values ​​(i.e., the first learning result), calculates, through numerical computation, one or more evaluation index values ​​related to processing time, processing accuracy, and surface quality for the processing target shape 320 or for each processing surface or each processing edge in the processing target shape 320. Depending on the balance between these values, within the determined range of command value generation parameters, the set of command value generation parameters that minimizes the evaluation index values ​​related to processing time, processing accuracy, and surface quality is called the command value generation parameter set candidate. In one example, the first optimal solution exploration unit 14 uses optimization algorithms such as grid search, random search, Newton's method, Bayesian optimization, or evolutionary computation to obtain the command value generation parameter set. An example of evolutionary computation is NSGA-II (Non-dominated Sorting Genetic Algorithms II), AGE-MOEA (Adaptive Geometry Estimation based a MultiObjective Evolutionary Algorithm), AGE-MOEA2, and R-NSGA-II (Reference point-based NSGA-II). The command value generation parameters are integrated into an exploratory command value generation parameter set. Furthermore, by inputting this exploratory command value generation parameter set into the first learning result, evaluation index values ​​related to processing time, processing accuracy, and surface quality are obtained, and the exploratory command value generation parameter set and the evaluation index values ​​are correlated. Among the combinations of exploratory command value generation parameters and corresponding evaluation index values, the command value generation parameter set corresponding to the best evaluation index value is selected as a candidate command value generation parameter set from the results classified using the evaluation index values ​​related to processing time, processing accuracy, and surface quality.

[0135] Figure 13 This is a diagram illustrating an example of a set of command value generation parameters related to the machining surface explored by the first optimal solution exploration unit in Implementation 1. Figure 13 The example shows a distribution plot of the evaluation index values ​​corresponding to the set of command value generation parameters, plotted on an orthogonal coordinate system with the evaluation index values ​​related to processing time, processing accuracy, and surface quality as axes. Additionally, this example shows the distribution plot of the evaluation index values ​​for processing time, processing accuracy, and surface quality, plotted on an orthogonal coordinate system with the evaluation index values ​​related to processing time, processing accuracy, and surface quality as axes. Figures 2 to 4 This is an example of the result obtained by exploring the parameter set generated from the instruction values ​​of the machining surface S1 with a machining target shape of 320. Figure 13The extracted command value generation parameter set candidates can be categorized into four types: a priority mode (processing time priority) that shortens processing time among the three evaluation indicators (processing time, processing accuracy, and surface quality); a priority mode (processing accuracy priority) that improves processing accuracy; a priority mode (surface quality priority) that improves surface quality; and a balanced mode that improves the balance of the three evaluation indicators. Evaluation indicator values ​​outside these four modes are related to other command value generation parameter sets.

[0136] exist Figure 13 The document records one example extracted from each of the four categories of command value generation parameter sets: processing time priority mode, processing accuracy priority mode, surface quality priority mode, and balance mode. However, it is not necessary to extract command value generation parameter sets corresponding to all categories; extracting those corresponding to at least one category is sufficient. Alternatively, multiple command value generation parameter sets corresponding to a single category can be extracted.

[0137] As described above, the first optimal solution exploration unit 14 has the following function: it explores candidate instruction value generation parameter sets from a combination of the instruction value generation parameter set for exploration and the evaluation index values ​​obtained when the instruction value generation parameter set for exploration is input to the first learning result. These candidate instruction value generation parameter sets include either one of the following: an instruction value generation parameter set that prioritizes improving any one of the evaluation index values ​​for processing time, processing accuracy, and surface quality within the determined instruction value generation parameter range, and another instruction value generation parameter set that improves the evaluation index values ​​for processing time, processing accuracy, and surface quality in a balanced manner within the determined instruction value generation parameter range. When processing the entire workpiece under one condition, the candidate instruction value generation parameter set explores the processing target shape 320. When the entire workpiece is divided into multiple parts, and each separate part is processed under different conditions, the candidate instruction value generation parameter set explores each processing surface or each processing edge. Furthermore, the first optimal solution exploration unit 14 can also perform learning processing and inference processing simultaneously.

[0138] Return to Figure 1The candidate information storage unit 15 stores candidate information, which is the information that associates the command value generation parameter set candidates extracted by the first optimal solution exploration unit 14 with the evaluation index values ​​and the processing feature quantities calculated by the feature quantity calculation unit 11. In one example, when multiple command value generation parameter set candidates are extracted, the multiple command value generation parameter set candidates are associated with each evaluation index value and the processing feature quantities calculated by the feature quantity calculation unit 11 and stored in the candidate information storage unit 15. When the entire workpiece is processed under one condition, the candidate information is obtained by associating the command value generation parameter set candidates with the evaluation index values ​​and the processing feature quantities for each processing target shape 320. In addition, when the entire workpiece is divided into multiple parts and each separate part is processed under different conditions, the candidate information is obtained by associating the command value generation parameter set candidates with the evaluation index values ​​and the processing feature quantities for each processing surface or each processing edge. Furthermore, the evaluation index values ​​generated from all instruction values ​​obtained through the learning and exploration processes in the first optimal solution exploration unit 14, as well as the processed feature values ​​calculated by the feature quantity calculation unit 11, can be stored in the candidate information storage unit 15.

[0139] The preference information setting unit 16 performs the following control: it sets the command value generation parameter set candidates to the command value generation device 3 that generates tool movement commands, and displays them on the display unit 17 in association with the machining feature quantities and various evaluation index values ​​calculated when the command value generation device 3 operates. When machining the entire workpiece under one condition, the preference information setting unit 16 displays the machining feature quantities and various evaluation index values ​​of the command value generation parameter set candidates on the display unit 17 in association with the machining target shape 320. Furthermore, when the entire workpiece is divided into multiple parts, and each part is machined under different conditions, the preference information setting unit 16 associates the machining feature quantities and various evaluation index values ​​of the command value generation parameter set candidates, and displays them on the display unit 17 for each machining surface or each machining edge. Alternatively, the actual machine tool can be operated according to the command value generation parameter set, and the operator can be prompted with the actual machined target shape 320 of the workpiece in association with the various evaluation index values. Image data of the machined target shape 320 of the workpiece can also be displayed on the display unit 17 in association with the various evaluation index values.

[0140] Furthermore, the preference information setting unit 16 sets preference information for the evaluation index values ​​of the parameter set candidates generated by the operator among the processing feature quantities and various evaluation index values ​​displayed on the display unit 17. In one example, the preference information setting unit 16 sets the evaluation index values ​​of the parameter set candidates generated by the operator among the processing feature quantities and various evaluation index values ​​displayed on the display unit 17, after the operator has selected and adjusted the command values, as preference information. When processing the entire workpiece under one condition, the preference information is set for the processing target shape 320. When the entire workpiece is divided into multiple parts, and each separate part is processed under different conditions, the preference information is set for each processing surface or each processing edge of the processing target shape 320. In addition, the display control unit corresponds to the preference information setting unit 16.

[0141] Furthermore, during the processing performed by the first optimal solution exploration unit 14, preference information can be preset within the scope that the operator can perform. At this time, in the first optimal solution exploration unit 14, candidate parameter sets for generating instruction values ​​that reflect the preset preference information are extracted. Therefore, the preference information setting unit 16 can set preference information other than the preset items, and the preference information setting unit 16 can also reset the preset items.

[0142] In one example, the preference information setting unit 16 displays the instruction value generation parameter set candidates stored in the candidate information storage unit 15, along with the evaluation index values ​​associated with the instruction value generation parameter set candidates and the processing feature quantities on the display unit 17. In this example, the instruction value generation parameter set candidates are processing time priority mode, processing accuracy priority mode, surface quality priority mode, and balance mode. Based on the processing feature quantities obtained by the feature quantity calculation unit 11 via the input unit (not shown), the operator selects one instruction value generation parameter set candidate from the four categories: processing time priority mode, processing accuracy priority mode, surface quality priority mode, and balance mode. Furthermore, the operator sets their preference information via the input unit, using the processing feature quantities obtained from the selected instruction value generation parameter set candidate and the evaluation index values ​​related to processing time, processing accuracy, and surface quality as references. The preference information consists of the evaluation index values ​​set by the operator, i.e., evaluation index values ​​related to the operator's processing time, processing accuracy, and surface quality. The preference information can be described as information indicating which aspect of processing time, processing accuracy, and surface quality the operator prioritizes during processing. When machining the entire workpiece under a single condition, the operator sets preference information regarding machining time, machining accuracy, and surface quality for the target shape 320. Alternatively, when the workpiece is divided into multiple parts and each part is machined under different conditions, the operator sets preference information regarding machining time, machining accuracy, and surface quality for each machined surface or each machined edge. In the latter case, the operator adjusts the evaluation index values ​​for machining time, machining accuracy, and surface quality related to the selected parameter set candidate for the target machined surface or edge. This adjustment is based on the operator's preferences. The preference information setting unit 16 sets the adjusted evaluation index values ​​for machining time, machining accuracy, and surface quality as preference information for the target machined surface or edge.

[0143] exist Figures 2 to 4 In the example shown, the machining target shape 320 is composed of machining surfaces S1-S3 and machining edges E1 and E2. The operator selects the parameter set candidate from the instruction values ​​that are closest to the operator's preferences in each of the machining surfaces S1-S3 and machining edges E1 and E2. Figure 14 It means about from Figure 13 This diagram illustrates an example of setting preference information by the operator to generate a parameter set candidate by selecting one instruction value from the categories of machining time priority mode, machining accuracy priority mode, surface quality priority mode, and balance mode related to the machining surface. Figure 14 also with Figure 13 Similarly, the evaluation index values ​​related to the machined surface S1 are shown. For example... Figure 14As shown, as a specific process, if the operator pre-indicates the position on the machining surface S1 of the target shape 320, the current machining time, machining accuracy, and surface quality evaluation index values ​​related to the indicated machining surface S1 are displayed on the display unit 17. In one example, the machining time, machining accuracy, and surface quality evaluation index values ​​associated with the parameter set candidate generated by the instruction value selected by the operator are displayed. The operator corrects the displayed current machining time, machining accuracy, and surface quality evaluation index values ​​via the input unit. The preference information setting unit 16 sets the corrected machining time, machining accuracy, and surface quality evaluation index values ​​as preference information on the machining surface S1. Figure 14 In the example, the noodle quality priority mode was selected by the operator, and the method of shortening processing time while maintaining noodle quality was adjusted.

[0144] In this example, the method for specifying the machining surface involves using a pointing device such as a mouse or touch panel, allowing the operator to select a position on the machining surface of the target shape 320. Alternatively, the indicated position can be a specific point, multiple points, or a continuous area.

[0145] In another example, the method for correcting the evaluation index value can be numerical input or continuous adjustment of the current setting value using GUI (Graphical User Interface) buttons such as buttons or bars. In this case, the maximum and minimum values ​​of the evaluation index value corresponding to the parameter set candidates can be generated based on the instruction values ​​stored in the candidate information storage unit 15 of the parameter adjustment device 1, and the range that can be input or adjusted can be set.

[0146] Furthermore, the preference information setting unit 16 can generate evaluation index values ​​and processing feature quantities corresponding to the parameter set candidates generated from the instruction values ​​stored in the candidate information storage unit 15 of the parameter adjustment device 1, and predict the processing feature quantities obtained when preference information is set, and display them in association with the processing target shape 320 on the display unit 17, etc. In one example, consider the following method: linearly interpolate the processing feature quantity of the evaluation index value that is closest to the set preference information from the evaluation index values ​​generated from the instruction values ​​stored in the candidate information storage unit 15 of the parameter adjustment device 1, and the processing feature quantity of the evaluation index value before the preference information is set, and predict the processing feature quantity for the set preference information.

[0147] Preference information can be set for all processing time, processing accuracy, and surface quality, or it can be set for only a portion of them. Furthermore, if preference information is not set by the operator, the preference information setting unit 16 interprets it as the same as the current evaluation index value being set as selection information and sets the preference information accordingly.

[0148] Return to Figure 1 The display unit 17 displays the stored information in the candidate information storage unit 15 according to the instructions from the preference information setting unit 16. In one example, the display unit 17 displays the processing feature quantities of the command value generation parameter set candidates and the respective evaluation index values ​​in association. When the entire workpiece is processed under one condition, the display unit 17 displays the processing feature quantities of the command value generation parameter set candidates and the respective evaluation index values ​​in association for the processing target shape 320. In addition, when the entire workpiece is divided into multiple parts and each separate part is processed under different conditions, the display unit 17 associates the processing feature quantities of the command value generation parameter set candidates and the respective evaluation index values, and displays them for each processing surface or each processing edge of the processing target shape 320.

[0149] The second optimal solution exploration unit 18 explores the instruction value generation parameter set corresponding to the evaluation index value that minimizes the difference with the preference information. That is, the second optimal solution exploration unit 18 explores one instruction value generation parameter set from multiple instruction value generation parameter sets, in a manner where the evaluation index value is close to the preference information set in the preference information setting unit 16. Specifically, the second optimal solution exploration unit 18 repeatedly performs the action of obtaining the difference between the evaluation index value (which evaluates processing time, processing accuracy, and surface quality corresponding to the instruction value generation parameter set) and the operator's preference information regarding processing time, processing accuracy, and surface quality, and calculates the instruction value generation parameter set that minimizes the difference between the evaluation index value and the operator's preference information. The calculated instruction value generation parameter set can be one or multiple. When processing the entire workpiece under one condition, the second optimal solution exploration unit 18 calculates the instruction value generation parameter set that minimizes the difference between the evaluation index value and the operator's preference information for the processing target shape 320. Furthermore, when the workpiece is divided into multiple parts and each part is processed under different conditions, the second optimal solution exploration unit 18 generates a parameter set by finding the instruction value that minimizes the difference between the evaluation index value and the operator's preference information for each processing surface and each processing edge in the processing target shape 320.

[0150] The second optimal solution exploration unit 18 uses the instruction value generation parameter set, the evaluation index value corresponding to the instruction value generation parameter set, and the difference between the evaluation index value and the operator's preference information as learning data, and performs learning processing using a neural network. If the relationship between the instruction value generation parameters and the difference between the evaluation index value and preference information can be obtained, other methods besides neural networks can also be used to learn the relationship between the instruction value generation parameter set and the difference between the evaluation index value and preference information. In one example, to obtain the relationship between the instruction value generation parameter set and the difference between the evaluation index value and preference information, a simple function such as a quadratic polynomial can be used, or a probabilistic model such as a Gaussian process model can be used. As described above, the second optimal solution exploration unit 18 uses the learning data, including the instruction value generation parameter set and the difference between the evaluation index value and preference information corresponding to the instruction value generation parameter set, to generate a second learning result for inferring the difference between the evaluation index value and preference information corresponding to the instruction value generation parameter set based on the instruction value generation parameter set.

[0151] In addition, the difference between the evaluation index value and the operator's preference information is a 3-dimensional data of processing time, processing accuracy and surface quality, but it can also be transformed into 1-dimensional data such as norm and used as learning data.

[0152] Furthermore, the second learning result in the second optimal solution exploration unit 18 can be the first learning result obtained based on the learning process of the first optimal solution exploration unit 14, or it can be the learning result obtained by additionally implementing the learning process on the first learning result obtained based on the learning process of the first optimal solution exploration unit 14.

[0153] The second optimal solution exploration unit 18, based on the relationship between the instruction value generation parameter set (a learning result) and the difference between the evaluation index value and preference information, calculates, through numerical computation, the instruction value generation parameter set that minimizes the difference between the evaluation index value and the operator's preference information regarding processing time, processing accuracy, and surface quality. In other words, the second optimal solution exploration unit 18 uses the relationship—the second learning result—used to infer the difference between the evaluation index value and preference information corresponding to the instruction value generation parameter set, to infer the difference between the evaluation index value and preference information corresponding to the exploration instruction value generation parameter set, and uses the inferred result to explore the instruction value generation parameter set that minimizes the difference between the evaluation index value and preference information. In one example, the second optimal solution exploration unit 18 uses optimization algorithms such as grid search, random search, Newton's method, Bayesian optimization, or evolutionary computation to find the exploration instruction value generation parameter set. An example of evolutionary computation is NSGA-II, AGE-MOEA, AGE-MOEA2, and R-NSGA-II.

[0154] Furthermore, the second optimal solution exploration unit 18 calculates the difference between the evaluation index value obtained by inputting the calculated instruction value generation parameter set for exploration into the relational expression and the operator's preference information. It also calculates the instruction value generation parameter set that minimizes the difference between the evaluation index value and the preference information. Preferably, it calculates one instruction value generation parameter set that minimizes the difference between the evaluation index value and the preference information. However, it is also possible to calculate multiple instruction value generation parameter sets sequentially from the instruction value generation parameter set that minimizes the difference between the evaluation index value and the preference information, until all instruction value generation parameter sets are calculated where the difference between the evaluation index value and the preference information converges to within a threshold set by the operator. That is, the second optimal solution exploration unit 18 can explore instruction value generation parameter sets corresponding to evaluation index values ​​whose difference with the preference information converges to within a certain value. In this case, the instruction value generation parameter set explored by the second optimal solution exploration unit 18 may be one or multiple. The instruction value generation parameter set calculated in the above manner is called the adjusted instruction value generation parameter set. The second optimal solution exploration unit 18 stores the calculated adjusted instruction value generation parameter set in the adjusted instruction value generation parameter set storage unit 19. The instruction value generation parameter set used by the second optimal solution exploration unit 18 for exploration corresponds to the instruction value generation parameter set used for the second exploration. The second optimal solution exploration unit 18 can also perform learning processing and inference processing simultaneously.

[0155] The adjusted command value generation parameter set storage unit 19 stores the adjusted command value generation parameter set explored by the second optimal solution exploration unit 18. When machining the entire workpiece under one condition, the adjusted command value generation parameter set storage unit 19 stores the adjusted command value generation parameter set calculated for the machining target shape 320. Furthermore, when the entire workpiece is divided into multiple parts, and each part is machined under different conditions, the adjusted command value generation parameter set storage unit 19 stores the adjusted command value generation parameter set calculated for each machining surface and each machining edge in the machining target shape 320. The command value generation device 3 rewrites the set value of the command value generation parameter set to the adjusted command value generation parameter set extracted by the second optimal solution exploration unit 18. Moreover, by using the set command value generation parameters to operate the command value generation device 3 and perform machining, a machining result matching the operator's preferences can be obtained.

[0156] Next, the parameter adjustment method in the parameter adjustment device 1 of the structure described above will be explained. Figure 15 This is a flowchart illustrating an example of the sequence of parameter adjustment methods involved in Embodiment 1. Furthermore, here, an example is given of dividing the entire workpiece into multiple parts, and processing each separate part under different conditions.

[0157] First, the parameter adjustment device 1 and the command value generation device 3 are initially set (step S11). Specifically, the shape of the target object to be machined, namely the machining target shape 320, including the surface to be machined, is externally input to the parameter adjustment device 1. In addition, in the command value generation device 3, a machining program 310, which describes the tool path movement command corresponding to the machining target shape 320 and the movement speed command at that time, is externally input.

[0158] Next, the instruction value generation device 3 outputs the tool movement instruction per unit time according to the externally input machining program 310 (step S12). Then, the feature quantity calculation unit 11 simulates the movement of the controlled object, i.e., the working machine, on the computer, and estimates the actual tool tip point based on the output of the instruction value generation device 3, i.e., the interpolation point (step S13).

[0159] Next, the feature quantity calculation unit 11 calculates the machining feature quantity at each tool tip point estimated in step S13 in association with the machining surface or machining edge of the machining target shape 320 (step S14). Then, the evaluation index calculation unit 12 calculates the evaluation index values ​​for each machining time, machining accuracy, and surface quality based on the machining feature quantity calculated in step S14 (step S15).

[0160] Next, the first optimal solution exploration unit 14 takes the evaluation index values ​​and parameter ranges related to processing time, processing accuracy, and surface quality as input, learns the relationship between the command value generation parameter set and the evaluation index values ​​calculated by the evaluation index calculation unit 12, and outputs the first learning result (step S16). Then, based on the relationship between the command value generation parameters and the evaluation index values ​​as the first learning result, the first optimal solution exploration unit 14 calculates the balance differences of the evaluation index values ​​related to processing time, processing accuracy, and surface quality for each processing surface or each processing edge in the processing target shape 320 through numerical calculation, and generates command value generation parameter set candidates that simultaneously optimize the evaluation index values ​​related to processing time, processing accuracy, and surface quality (step S17). In one example, for the processing surface or processing edge, four command value generation parameter set candidates are obtained respectively: processing time priority mode, processing accuracy priority mode, surface quality priority mode, and balance mode. In addition, one command value generation parameter set candidate can be obtained for each processing surface or each processing edge, or multiple can be obtained.

[0161] Then, the preference information setting unit 16 generates a set of candidate instruction values ​​for the processed feature quantities, and displays the processed feature quantities, processing time, processing accuracy, and surface quality evaluation index values ​​of the candidate instruction value generation parameter sets on the display unit 17 for each processed surface and each processed edge (step S18). That is, the preference information setting unit 16 displays the processed feature quantities and various evaluation index values ​​calculated when the instruction value generation device 3 is activated, which are set as candidate instruction value generation parameter sets, on the display unit 17 in association. As described above, the processed feature quantities and various evaluation index values ​​related to the candidate instruction value generation parameter sets are displayed to the operator in association. Then, the operator can select the processed feature quantity, i.e., the instruction value generation parameter set, corresponding to the evaluation index value that matches or is close to the operator's preference from the displayed content. As a result, the instruction value generation parameters can be made to converge to the operator's preference.

[0162] The operator selects a set of candidate parameters for each machining surface and each machining edge based on the machining feature values, and adjusts the evaluation index values ​​of machining time, machining accuracy, and surface quality as needed. The preference information setting unit 16 uses the adjusted machining time, machining accuracy, and surface quality evaluation index values ​​as preference information and sets them for each machining surface or each machining edge (step S19).

[0163] Next, based on the second learning result, the second optimal solution exploration unit 18 repeatedly performs the action of obtaining the difference between the evaluation index value for evaluating the processing time, processing accuracy and surface quality corresponding to the instruction value generation parameter set and the operator's preference information regarding processing time, processing accuracy and surface quality. For each processing surface or each processing edge in the processing target shape 320, the adjusted instruction value generation parameter set that minimizes the difference between the evaluation index value and the operator's preference information is obtained (step S20).

[0164] Then, the instruction value generation device 3 rewrites the set value of the instruction value generation parameter set with the adjusted instruction value parameter set extracted in the second optimal solution exploration unit 18 and performs processing, thereby obtaining a processing result that matches the operator's preferences. Furthermore, if the operator's preferences change, by restarting the processing steps S18 to S20 that reflect the operator's preferences, the optimal adjusted instruction value generation parameter set can be calculated for operators whose preferences have changed in a short period of time. This concludes the parameter adjustment method. While an overview of each step has been provided here, the details of each step are as described above.

[0165] Furthermore, when machining the entire workpiece under one condition, in step S14, the feature quantity calculation unit 11 calculates the machining feature quantity at the tool tip point in association with the machining target shape 320. In step S17, the first optimal solution exploration unit 14 obtains a parameter set candidate for generating command values ​​for the machining target shape 320. In step S18, the preference information setting unit 16 displays the machining feature quantity, machining time, machining accuracy, and surface quality evaluation index values ​​of the parameter set candidate for generating command values ​​related to the machining target shape 320 on the display unit 17. In addition, in step S19, the preference information setting unit 16 sets the machining time, machining accuracy, and surface quality evaluation index values ​​adjusted by the operator as preference information for the machining target shape 320. Moreover, in step S20, the second optimal solution exploration unit 18 obtains an adjusted parameter set for generating command values ​​related to the machining target shape 320.

[0166] As described above, in Embodiment 1, a first learning result is used to infer an evaluation index value greater than or equal to one for evaluating the processing result based on the instruction value generation parameter set. This infers the evaluation index value corresponding to the explored instruction value generation parameter set, and explores multiple instruction value generation parameter set candidates that simultaneously optimize each evaluation index value using the inferred result. Furthermore, the processing feature quantity calculated when the explored instruction value generation parameter set candidate is set on the instruction value generation device 3 and operated is displayed on the display unit 17 in association with each evaluation index value. Thus, the operator can observe the evaluation index value and processing feature quantity related to the multiple instruction value generation parameter set candidates, and select candidates that match or are close to the operator's preferences as the instruction value generation parameter set. Moreover, by using the selected instruction value generation parameter set, the instruction value generation parameter set matching the operator's preferences can converge faster than before. In other words, an environment that allows the instruction value generation parameter set matching the operator's preferences to converge faster than before can be provided to the operator.

[0167] Furthermore, through Implementation Method 1, processing time, processing accuracy, and surface quality can be used as evaluation indicators, and parameters can be automatically adjusted to match the processing target shape 320 with the operator's preferences. Thus, it is possible to achieve the desired processing accuracy while minimizing processing time. Specifically, it has the effect that the parameter set for generating instruction values ​​that match the operator's preferences converges more quickly than before.

[0168] Furthermore, in cases where adjustments that match the operator's preferences cannot be obtained, or where the operator's preferences have changed, the execution should proceed from... Figure 15The processing up to steps S18 to S20 is sufficient. Thus, the operator can perform fine-tuning and correction of the instruction value generation parameter set with minimal workload and time.

[0169] Furthermore, in the technology described in Patent Document 1, the action is confirmed through multiple parameter sets, and the most appropriate parameter set can be selected. However, a single parameter set is applied through a series of processes. That is, in the technology described in Patent Document 1, it is not possible to locally set the optimal parameters. Figure 16 This diagram illustrates an example of the machining process for a component with a blade shape. (Example:) Figure 16 As shown, in one example, when machining a blade-shaped component 400 using a cutting tool 40, it is impossible to address the desire for high-precision machining of the rounded end portions 401 and high-speed machining of the flat portion 402 between the end portions 401 using a single set of parameters. Furthermore, if the test program used for parameter adjustment differs from the shape of the workpiece being machined, the accuracy of the parameter adjustment deteriorates, failing to obtain an adjustment result that matches the operator's preferences. However, in Embodiment 1, when the workpiece is divided into multiple parts and each part is machined under different conditions, a parameter set is generated by calculating adjusted command values ​​reflecting the operator's preferences for each machined surface and each machined edge of the target shape 320. Figure 16 In the example, the adjusted command value generation parameters, reflecting the different operator preferences in the two-end portion 401 and the flat portion 402, are obtained. Thus, the set of command value generation parameters corresponding to the evaluation index values ​​matching the operator's preferences can converge faster than before, while considering the overall or partial shape of the workpiece. Furthermore, it has the effect that each part of a workpiece can be processed according to the operator's preferences.

[0170] Furthermore, the learning data used in the second optimal solution exploration unit 18 can be data obtained from the same control object as the control object that obtained the learning data used in the first optimal solution exploration unit 14. Alternatively, the learning data used in the second optimal solution exploration unit 18 can be data obtained from a different control object than the control object that obtained the learning data used in the first optimal solution exploration unit 14. That is, the learning results of the first optimal solution exploration unit 14 and the second optimal solution exploration unit 18 in Embodiment 1 can be learning results obtained in different control objects. In one example, the learning result of the second optimal solution exploration unit 18 can be a learning result obtained in an actual machine, and the learning result of the first optimal solution exploration unit 14 can be a learning result obtained through simulation of the machine's movements on a computer. By adopting the structure described above, when changes in the machine's state, such as historical changes or thermal displacement, occur, after matching the operator's preferences to a certain extent in the simulation, high-precision adjustments can be made in the actual machine with fewer adjustments, thereby enabling automatic parameter adjustments that match the operator's preferences.

[0171] Implementation method 2.

[0172] Figure 17 This is a diagram illustrating an example of the structure of the parameter adjustment device according to Embodiment 2. Furthermore, structural elements identical to those in Embodiment 1 are labeled with the same reference numerals and their descriptions are omitted; only the parts different from those in Embodiment 1 are described. In the structure of Embodiment 1, the parameter adjustment device 1A further includes a shape analysis unit 20 that analyzes shape information—information representing the shape of each processing surface or each processing edge of the processing target shape 320—based on processing feature quantities.

[0173] The shape analysis unit 20 analyzes shape information, i.e., shape information, representing the shape of each machining surface or each machining edge of the machining target shape 320, based on the machining feature quantities calculated by the feature quantity calculation unit 11. In one example, the shape analysis unit 20 extracts adjacent paths, i.e., adjacent paths, which are tool tip point paths representing each machining surface or each machining edge of the machining target shape 320, and derives shape information based on the machining feature quantities corresponding to the extracted adjacent paths. In one example, the shape analysis unit 20 uses the cumulative value of the tangent vector change calculated based on the machining feature quantities corresponding to the adjacent paths as shape information. Alternatively, the shape analysis unit 20 may use the average value of the tangent vector change derived from the machining feature quantities as shape information. In this case, the machining feature quantity is the velocity of the tool tip point. In other examples, the shape analysis unit 20 uses a function fitted by one-dimensionalizing the distance from the centroid of the adjacent paths as shape information. In this case, the shape information can also be obtained by fitting a simple function such as a quadratic polynomial.

[0174] Furthermore, as a method for extracting adjacent paths, in one example, there are processing features associated with each processing surface or each processing edge in the processing target shape 320. Therefore, adjacent paths are extracted by grouping each feature of continuous processing in the time series.

[0175] The shape analysis unit 20 stores the shape information obtained based on the adjacent paths calculated in the above manner in the evaluation index information storage unit 13, associating it with the instruction value generation parameter set and the evaluation index values ​​related to processing time, processing accuracy, and surface quality. As described above, in Embodiment 2, the evaluation index information associates the evaluation index values, instruction value generation parameter set, and shape information related to processing time, processing accuracy, and surface quality with respect to the processing surface or processing edge of the processing target shape 320. Furthermore, the evaluation index information may also have corresponding processing feature quantities in addition to the evaluation index values, instruction value generation parameter set, and shape information related to processing time, processing accuracy, and surface quality.

[0176] The first optimal solution exploration unit 14 learns the relationship between the parameter set in the command value generation device 3 (i.e., the command value generation parameter set) and the evaluation index values ​​calculated by the evaluation index calculation unit 12, by adding shape information derived from the shape analysis unit 20, and obtains a first learning result. Using the first learning result, the first optimal solution exploration unit 14 explores one or more command value generation parameter sets that simultaneously optimize the evaluation index values ​​of processing time, processing accuracy, and surface quality. When exploring multiple command value generation parameter sets, the balance of evaluation index values ​​in a compromise relationship becomes different, and the unit explores command value generation parameter sets that simultaneously optimize the evaluation index values ​​of processing time, processing accuracy, and surface quality.

[0177] Here, the learning process of the first optimal solution exploration unit 14 will be explained. The first optimal solution exploration unit 14 takes as input the evaluation index values, instruction value generation parameter set and shape information related to processing time, processing accuracy and surface quality, and the instruction value generation parameter set, the evaluation index values ​​calculated by the evaluation index calculation unit 12 and the shape information, and outputs the first learning result.

[0178] Specifically, a neural network is constructed that takes the parameter set of instruction values ​​and shape information as input and the evaluation index value as output. The first optimal solution exploration unit 14 learns by updating the weighting coefficients of the neural network.

[0179] The first optimal solution exploration unit 14 selects and outputs a set of command value generation parameters for performing the next processing action from a predetermined parameter range within the processing target shape 320. When selecting the next set of command value generation parameters, the first optimal solution exploration unit 14 can select a set of command value generation parameters representing excellent evaluation index values ​​based on the first learning result, or it can sequentially select each set of command value generation parameters from points of an equally spaced grid. The first optimal solution exploration unit 14 has the function of updating the function for calculating evaluation index values ​​related to processing time, processing accuracy, and surface quality based on the command value generation parameter set and shape information.

[0180] The first optimal solution exploration unit 14 repeatedly performs the action of obtaining the evaluation index value corresponding to the instruction value generation parameter set and shape information. The first optimal solution exploration unit 14 uses the instruction value generation parameter set and the evaluation index value and shape information corresponding to the instruction value generation parameter set as learning data, and performs learning processing using a neural network as described in Embodiment 1.

[0181] Through the above actions, the first learning result is obtained as a function that takes the parameter set and shape information generated by the neural network relation (i.e., instruction value) as input and the evaluation index value as output.

[0182] If the first learning result is used, even without generating a parameter set and shape information for the new instruction value and performing the machining action, it is possible to obtain the evaluation index values ​​Qt, Qa, and Qq for the machining time, machining accuracy, and surface quality corresponding to the parameter set and shape information generated by the new instruction value.

[0183] Furthermore, in Implementation 2, a neural network was used to construct the relationship between the instruction value generation parameter set, shape information, and evaluation index value. However, if the relationship between the instruction value generation parameter set, shape information, and evaluation index value can be obtained, methods other than neural networks can be used. In one example, to obtain the relationship between the instruction value generation parameter set, shape information, and evaluation index value, a simple function such as a quadratic polynomial or a probabilistic model such as a Gaussian process model can be used.

[0184] Next, the parameter adjustment method in Implementation Method 2 will be explained. Figure 18 This is a flowchart illustrating an example of the sequence of parameter adjustment methods involved in Embodiment 2. Furthermore, regarding Embodiment 1... Figure 15 The same process is labeled with the same step number, but its description is omitted.

[0185] In Implementation Method 2, after step S15, the shape analysis unit 20 analyzes the shape information for each processed surface or each processed edge of the processing target shape 320 based on the processing feature quantities calculated in step S14 (step S31). Next, replacing the processing in step S16, the first optimal solution exploration unit 14 takes the evaluation index values ​​related to processing time, processing accuracy, and surface quality, the command value generation parameter set, and the shape information as input, and learns the relationship between the command value generation parameter set, the evaluation index values ​​calculated by the evaluation index calculation unit 12, and the shape information, and outputs the first learning result (step S32). Then, the process jumps to step S17.

[0186] As described above, according to Embodiment 2, learning is performed by adding shape information of the processed surface or processed edge compared to Embodiment 1, resulting in a first learning result. Therefore, even if the processing target shape 320 desired by the operator is changed, the three evaluation indicators—processing time, processing accuracy, and surface quality—can be used to automatically adjust the parameters for each processed surface or each processed edge in the processing target shape 320, matching the operator's preferences.

[0187] Furthermore, the first learning result of the first optimal solution exploration unit 14 and the second learning result of the second optimal solution exploration unit 18 in Embodiment 2 can be obtained from different controlled objects, similar to Embodiment 1. In one example, the second optimal solution exploration unit 18 may use the second learning result, which uses learning data obtained in the actual machine, while the first optimal solution exploration unit 14 may use the first learning result, which uses learning data obtained in a simulation of the machine's movements on a computer. By adopting the structure described above, in the event of changes in the machine's historical state or thermal displacement, after matching the operator's preferences to a certain extent in the simulation, adjustments can be made with high precision in the actual machine with fewer adjustments, automatically adjusting the command value generation parameter set to match the operator's preferences. Of course, the first learning result of the first optimal solution exploration unit 14 and the second learning result of the second optimal solution exploration unit 18 can be obtained from the same controlled object.

[0188] Furthermore, the first learning result of the first optimal solution exploration unit 14 and the second learning result of the second optimal solution exploration unit 18 in Embodiment 2 can be the first learning result and the second learning result obtained in various different processing procedures 310. In one example, the first learning result can be obtained in the first optimal solution exploration unit 14 using a general processing procedure 310 that can correspond to various shape information, and the second learning result can be obtained in the second optimal solution exploration unit 18 using the processing procedure 310 in the processing target shape 320 desired by the operator. Thus, when the operator changes the desired processing target shape 320, the operator can automatically adjust the instruction value generation parameter set with less workload and time.

[0189] Next, the hardware structure of the parameter adjustment devices 1 and 1A will be described. In embodiments 1 and 2, the parameter adjustment devices 1 and 1A execute computer programs, i.e., programs, on a computer system, thereby enabling the computer system to function as the parameter adjustment devices 1 and 1A.

[0190] Figure 19 This is a diagram illustrating an example of the structure of a computer system that implements the parameter adjustment device according to embodiments 1 and 2. For example... Figure 19 As shown, the computer system has a control unit 901, an input unit 902, a storage unit 903, a display unit 904, a communication unit 905, and an output unit 906, which are connected via a system bus 907.

[0191] exist Figure 19In this system, the control unit 901, in one example, is a processor such as a CPU (Central Processing Unit), which executes the program describing the processing in the parameter adjustment devices 1 and 1A of embodiments 1 and 2. The input unit 902, in one example, is composed of a keyboard, mouse, etc., for users of the computer system to input various information. The storage unit 903 includes various memory devices such as RAM (Random Access Memory), ROM (Read Only Memory), and hard disks, storing the program to be executed by the control unit 901, the data required during processing, etc. Additionally, the storage unit 903 is also used as a temporary storage area for programs. The display unit 904, composed of a monitor, liquid crystal display panel, etc., displays various screens to the user of the computer system. In one example, the input unit 902 and the display unit 904 can be configured as a touch panel integrally formed with the input unit 902 and the display unit 904. The communication unit 905 is a receiver and transmitter that performs communication processing. The output unit 906 is a printer, speaker, etc. Furthermore, Figure 19 This is one example; the structure of a computer system is not limited to... Figure 19 Examples.

[0192] Here, an example of the operation of the computer system up to the point where the program becomes executable will be described. In a computer system with the above structure, for example, a program is loaded into the storage unit 903 from a CD-ROM or DVD-ROM drive (not shown) mounted on a CD (Compact Disc)-ROM drive or a DVD (Digital Versatile Disc)-ROM drive. Furthermore, when the program is executed, the program read from the storage unit 903 is stored in the main storage area of ​​the storage unit 903. In this state, the control unit 901 executes the processing of the parameter adjustment devices 1 and 1A as described in Embodiments 1 and 2, according to the program stored in the storage unit 903.

[0193] Furthermore, in the above description, CD-ROM or DVD-ROM is used as the recording medium and a program describing the processing in parameter adjustment devices 1 and 1A is provided. However, it is not limited to this. Depending on the structure of the computer system and the capacity of the provided program, a program provided via a transmission medium such as the Internet through the communication unit 905 may also be used.

[0194] Figure 1 and Figure 17 The parameter adjustment device 1, the feature quantity calculation unit 11 of 1A, the evaluation index calculation unit 12, the first optimal solution exploration unit 14, the preference information setting unit 16, and the second optimal solution exploration unit 18 shown are as follows: Figure 17 The shape analysis unit 20 shown is obtained by... Figure 19The control unit 901 shown performs the following functions: Figure 19 The program stored in the storage unit 903 shown is used for implementation. It is also used in the implementation of the feature quantity calculation unit 11, the evaluation index calculation unit 12, the first optimal solution exploration unit 14, the preference information setting unit 16, the second optimal solution exploration unit 18, and the shape analysis unit 20. Figure 19 The storage unit 903 is shown. The evaluation index information storage unit 13, the candidate information storage unit 15, and the adjusted instruction value generation parameter set storage unit 19 are used... Figure 19 The storage unit 903 shown is used to implement this. The display unit 17 is implemented through... Figure 19 The display unit 904 shown is used to achieve this.

[0195] The structure shown in the above embodiments is an example, and it can also be combined with other known technologies, and the embodiments can be combined with each other. Without departing from the spirit of the subject, some parts of the structure can be omitted or changed.

[0196] Explanation of the label

[0197] 1. Parameter adjustment device, 3. Command value generation device, 11. Feature quantity calculation unit, 12. Evaluation index calculation unit, 13. Evaluation index information storage unit, 14. First optimal solution exploration unit, 15. Candidate information storage unit, 16. Preference information setting unit, 17. Display unit, 18. Second optimal solution exploration unit, 19. Adjusted command value generation parameter set storage unit, 20. Shape analysis unit, 310. Machining program, 320. Machining target shape, 321. Block, 321a. Upper surface, 322. Protrusion, E1, E2. Machining edges, S1, S2, S3. Machining curved surfaces.

Claims

1. A parameter adjustment device that adjusts a set of multiple parameters used to generate tool movement commands, namely a set of command value generation parameters, wherein the tool movement commands are composed of a set of interpolation points per unit time on the toolpath calculated based on a machining program used to machine a workpiece. The parameter adjustment device is characterized by having: The feature quantity calculation unit simulates the movement of the working machine of the controlled object according to the tool movement command and calculates the feature quantity of the machining. The evaluation index calculation unit calculates an evaluation index value that is greater than or equal to one based on the characteristic quantity of the processing. The first optimal solution exploration unit uses a first learning result obtained by learning the instruction value generation parameter set based on the instruction value generation parameter set and the evaluation index value to infer the evaluation index value. It then infers the evaluation index value corresponding to the instruction value generation parameter set used in the first exploration, and uses the inferred result to explore multiple instruction value generation parameter sets, i.e., instruction value generation parameter set candidates, that simultaneously optimize each of the evaluation index values. The display control unit sets the candidate instruction value generation parameter set to the instruction value generation device that generates the tool movement instruction, and displays the machining feature quantity and each of the evaluation index values ​​calculated when the instruction value generation device is activated on the display unit in association. The display control unit generates a set of parameter candidates for each of the evaluation index values ​​selected by the operator from among the processing feature quantities and various evaluation index values ​​displayed on the display unit, and sets preference information for each of the evaluation index values.

2. The parameter adjustment device according to claim 1, characterized in that, It also has a second optimal solution exploration unit, which explores the parameter set of instruction values ​​corresponding to the evaluation index value that minimizes the difference with the preference information.

3. The parameter adjustment device according to claim 1, characterized in that, It also has a second optimal solution exploration unit, which explores the parameter set of the instruction value corresponding to the evaluation index value whose difference with the preference information converges to a certain value.

4. The parameter adjustment device according to claim 2 or 3, characterized in that, The second optimal solution exploration unit uses a second learning result for inferring the difference between the evaluation index value and the preference information corresponding to the instruction value generation parameter set based on the instruction value generation parameter set. It then infers the difference between the evaluation index value and the preference information corresponding to the instruction value generation parameter set used in the second exploration, and uses the inferred result to explore the instruction value generation parameter set that minimizes the difference between the evaluation index value and the preference information.

5. The parameter adjustment device according to claim 4, characterized in that, The second optimal solution exploration unit has the following function: it generates the second learning result using learning data that includes the instruction value generation parameter set, the evaluation index value corresponding to the instruction value generation parameter set, and the difference between the preference information.

6. The parameter adjustment device according to claim 2 or 3, characterized in that, The first optimal solution exploration unit has the following function: it generates the first learning result using learning data that includes the parameter set generated by the instruction values ​​and the evaluation index values.

7. The parameter adjustment device according to claim 6, characterized in that, The learning data used in the second optimal solution exploration unit is obtained from the same control object as the learning data used in the first optimal solution exploration unit.

8. The parameter adjustment device according to claim 6, characterized in that, The learning data used in the second optimal solution exploration unit is data obtained from a control object that is different from the learning data used in the first optimal solution exploration unit.

9. The parameter adjustment device according to claim 1, characterized in that, The feature calculation unit calculates the processing feature quantity for each shape structural element of the target object, including the surface to be processed (i.e., the processing surface), which has one or more shape structural elements. The first optimal solution exploration unit explores candidate parameter sets for generating the instruction values ​​for each of the shape and structural elements. The display control unit sets the evaluation index values ​​of each candidate parameter set for each shape and structural element, based on the instruction values ​​selected and adjusted by the operator, as preference information.

10. The parameter adjustment device according to claim 1, characterized in that, The machining characteristic is the velocity of the tool tip. One of the evaluation index values ​​is a value related to processing time. The evaluation index value related to the machining time is the rate of deceleration of the speed of the tool tip relative to the commanded speed described in the machining program.

11. The parameter adjustment device according to claim 1, characterized in that, The machining characteristic quantity is the shape data of the target object to be machined, including the surface to be machined (i.e., the machined surface), and the distance between the target shape and the tool positioned at the tip of the tool (i.e., the machining error). One of the evaluation index values ​​is a value related to machining accuracy. The evaluation index value related to the machining accuracy is the average value of the machining error.

12. The parameter adjustment device according to claim 1, characterized in that, The machining characteristic quantity is the shape data of the target object to be machined, including the surface to be machined (i.e., the machined surface), and the distance between the target shape and the tool positioned at the tip of the tool (i.e., the machining error). One of the evaluation index values ​​is an evaluation index value related to surface quality. The evaluation index value related to the surface quality is the variance value of the processing error.

13. The parameter adjustment device according to claim 9, characterized in that, It also includes a shape analysis unit, which analyzes the shape information, i.e., shape information, of each shape structural element representing the shape of the processing target, based on the processing feature quantities calculated by the feature quantity calculation unit. The first optimal solution exploration unit learns by incorporating the shape information into the relationship between the instruction value generation parameter set and the evaluation index value, thereby generating the first learning result. The evaluation index value corresponding to the parameter set generated by the instruction value used in the first exploration and the shape information is inferred using the first learning result.

14. The parameter adjustment device according to claim 13, characterized in that, The shape analysis unit uses a function fitted by 1Dizing the cumulative value of the tangent vector change calculated based on the feature quantity of the machining corresponding to the adjacent path or the distance from the centroid of the adjacent path as the shape information. The adjacent path is the tool tip point path adjacent to the tool tip point path represented for each shape structural element.

15. A parameter adjustment method, comprising a parameter adjustment device, wherein the parameter adjustment device adjusts a set of multiple parameters used to generate tool movement commands, namely a set of command value generation parameters, wherein the tool movement commands are constituted by a set of interpolation points per unit time on the toolpath calculated based on a machining program used to machine an object. The characteristic of this parameter adjustment method is that it includes: The feature quantity calculation process involves simulating the movement of the working machine of the controlled object according to the tool movement command, and calculating the feature quantities to be processed. The evaluation index calculation process involves calculating an evaluation index value that is greater than or equal to one based on the characteristic quantity of the processing. The first optimal solution exploration step involves using a first learning result obtained from learning the instruction value generation parameter set based on the instruction value generation parameter set and the evaluation index value to infer the evaluation index value; inferring the evaluation index value corresponding to the instruction value generation parameter set used in the first exploration; and using the inferred result to explore multiple instruction value generation parameter sets, i.e., candidate instruction value generation parameter sets, that simultaneously optimize each of the evaluation index values; and The display control process involves setting the candidate set of command value generation parameters to the command value generation device that generates the tool movement command, and displaying the machining feature quantity and each of the evaluation index values ​​calculated when the command value generation device is activated in association on the display unit. The display control process sets preference information for each of the parameter set candidate evaluation index values ​​for the instruction values ​​selected by the operator among the processing feature quantities and various evaluation index values ​​displayed on the display unit.

Images