JP2007083359A - Grinding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a grinding method having high flexibility in calculating an appropriate shape to an actual processing shape and capable of satisfying accommodation capacity for a target processing shape and calculation time. <P>SOLUTION: This grinding method calculates a residence time during which a grinding tool for processing a workpiece stays at each coordinate, based on a shape of a workpiece, the target processing shape which is a difference between the shape of the workpiece and a design shape and a unit processing shape obtained in unit time processing; and grinds the workpiece with controlling the grinding tool to stay at each coordinate of the workpiece during the residence time; and includes a process for calculating a residence time distribution using fast Fourier transformation and a process for calculating the residence time distribution using a simulation of removing the residual shape. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a polishing method, and more particularly to a polishing method in which the surface of a hard and brittle material such as a semiconductor, glass, ceramics, a single metal or a single crystal of a metal oxide is precisely polished (aspherical surface polishing).

Conventionally, in high-precision optical elements typified by lenses, mirrors, etc. used for applications that require particularly high functions, polishing of the high-precision free-curved surface necessary for processing these is performed by shape measurement. This is done by repeating corrective polishing.
Such a shape correction of a workpiece is performed by positioning a small-diameter polishing tool on the surface of the workpiece so as to obtain a shape with a constant processing value processed per unit time (hereinafter referred to as a unit processing shape). The dwell time control is performed by controlling the speed and scanning.

The basic model is shown in FIG.
Here, a shape based on an error value (hereinafter referred to as an error shape) which is a difference shape between the surface shape of a workpiece obtained by shape measurement and a design shape is referred to as a shape based on a target machining value (hereinafter referred to as a target shape). Process shape d (x, y)).
Here, the shape machined by the residence time control by the unit machining shape f (x, y) and the residence time distribution g (x, y) calculated by convolution integration is referred to as a calculated machining shape.

で表される。
また、上記目標加工形状ｄ（ｘ，ｙ）から上記計算加工形状ｈ（ｘ，ｙ）を差し引いた形状を、残差形状と称する。
ここで、この計算加工形状は

で表される。 Here, this calculated shape is

It is represented by
A shape obtained by subtracting the calculated machining shape h (x, y) from the target machining shape d (x, y) is referred to as a residual shape.
Here, this calculated shape is

It is represented by

In order to perform the dwell time control, Equation 1 is developed, and the residual shape e (x, y) is minimized from the known unit machining shape f (x, y) and the target machining shape d (x, y). It is necessary to calculate the residence time distribution g (x, y).
However, this requires deconvolution and cannot be calculated with Equation 1 above. Therefore, conventionally, for example, a method using (A) fast Fourier transform as shown in FIG.
By the fast Fourier transform, the unit machining shape f (x, y) becomes F (w _x, w _y ), and the target machining shape d (x, y) becomes D (w _x, w _y ).
Also, the residence time distribution g (x, y) is G (w _x, w _y ), and the residual shape e (x, y) is E (w _x, w _y ).

に変換される。
また、これにより、上記式２はＥ（ｗ_x,ｗ_y）＝０として滞留時間分布を求める式

に展開することができる。 From the above, the above equation 1 is a simple function accumulation equation.

Is converted to
Accordingly, the above equation 2 is an equation for obtaining the residence time distribution with E (w _x, w _y ) = 0.

Can be deployed.

となる汎関数フィルタ

を用いて、滞留時間分布

を安定化させることができる。 Further, in the above formula 3, the solution is divergent in a region where the unit processing shape F (w _x, w _y ) approaches 0. So, in the whole frequency range

Functional filter

Using the residence time distribution

Can be stabilized.

Derivation of the actual residence time distribution is performed as shown in the flowchart of FIG.
First, (1) calculation of residence time distribution (determining residence time distribution by deconvolution from the unit machining shape and the target machining shape) is performed using the above equation 4.
Next, (2) the residence time distribution is corrected in accordance with restrictions on the processing machine.
Next, using Equation 2 above, (3) derivation of the residual shape (calculated machining shape is obtained by convolution integration from the corrected residence time distribution and unit machining shape, and the residual shape is calculated from the difference shape between the target machining shape and the calculated machining shape. Find the difference shape).
Next, (4) the residence time distribution is evaluated by determining whether the residual shape satisfies the required standard.
Next, if the residual shape does not satisfy the required standard, the residual shape is set as a new target machining shape (5) the target machining shape is reset, and the residual shape is requested again from (1). Repeat the above process until the standard is satisfied.
Finally, the required residual shape is obtained when the residual shape satisfies the standard. (6) The residence time distribution is derived.

Further, the calculation of the above (1) residence time distribution and (3) residual shape can also be performed by (B) a method using a removal simulation introduced as another method in Non-Patent Document 1 described above. it can.
In this removal simulation method, the unit time ΔT is determined, and the shape processed by the polishing tool in the unit time ΔT is defined as the unit processing shape, and (1) residence time distribution and (3) residual as described above are as follows. A shape is obtained.
First, (1) derivation of the residence time distribution is as follows. When a unit time polishing tool is scanned over the surface of the workpiece as shown in FIG. Subtract the unit processing shape from.
Next, a unit time ΔT is added to the corresponding location of the residence time distribution, and these are repeated until convergence, thereby obtaining a residence time distribution.
Further, the above (3) derivation of the residual shape is performed by superimposing the unit machining shapes for each residence time at each position on the workpiece surface as shown in FIG. Calculate the machining shape with. A residual shape is obtained by subtracting this calculated processing shape from the target processing shape.
Japan Society for Precision Engineering Vol.62 No.3 1996 p408-412

According to the method using (A) the fast Fourier transform described above, the calculation process can be simplified and various target machining shapes can be calculated in a relatively short time.
However, on the other hand, in order to perform the fast Fourier transform, there are various necessary constraints, so that the degree of freedom in calculating a shape close to the actual machining shape is low. Therefore, there is a portion that does not reflect the actual processing in the calculation of the above (1) residence time distribution and (3) residual shape, and the above (6) satisfactory evaluation of the residence time distribution. May not be performed.

On the other hand, the method using the (B) removal simulation described above has a high degree of freedom in calculating a shape close to the actual machining shape because the basic calculation process is simple. Therefore, it is possible to perform a calculation that accurately reflects actual processing.
However, in the above (1) residence time distribution, there is a target machining shape in which it is difficult to calculate the residence time distribution depending on the shape. In addition, the amount of calculation increases due to repetitive calculations, and in particular, the calculation of the above (1) residence time distribution takes an enormous amount of time.
Here, the advantages and disadvantages of (1) residence time distribution and (3) residual shape calculation by (A) the method using the fast Fourier transform and (B) the method using the removal simulation are summarized. As shown in FIG.
As is clear from this figure, in terms of the degree of freedom of calculation (a shape closer to the actual machining can be derived), both (1) residence time distribution and (3) residual shape calculation are It can be said that the method using simulation is excellent.

  Next, any of the above-described methods is superior in terms of the ability to cope with the target machining shape (the residence time distribution appropriate for various target machining shapes can be calculated). On the other hand, regarding the calculation of the (1) residence time distribution, the method using the (B) removal simulation is slightly inferior to the method using the (A) fast Fourier transform.

Finally, in terms of the time required for these calculations, in the calculation of (1) residence time distribution, the method using the (B) removal simulation is much larger than the method using the (A) fast Fourier transform. Often time is required.
Also, in the calculation of the residual shape (3), the method using the (B) removal simulation takes a longer time to calculate, but compared with the method (A) using the fast Fourier transform, It's not so different.

  In view of the above problems, the present invention has a high degree of freedom in calculating a shape close to the actual machining shape, and can satisfy the ability to cope with the target machining shape, the calculation time, and the like. Is intended to provide.

In order to solve the above problems, the present invention provides a polishing method configured as follows.
That is, the polishing method of the present invention has a shape of a workpiece, a target machining shape that is a difference between the shape of the workpiece and a design shape, and a unit machining shape that is a machining shape obtained when machining for a unit time. Based on this, a residence time for the polishing tool used for processing the workpiece to stay at each coordinate of the workpiece is calculated, and the polishing tool stays at each coordinate of the workpiece during the residence time. A polishing method for polishing a workpiece by controlling to a residence time distribution required to minimize a residual shape, which is an error shape remaining after processing, from the unit processing shape and the target processing shape, A calculation process obtained by superimposing the unit machining shape for the dwell time at each position on the workpiece surface and performing this over the entire machining area, with the step of calculating using fast Fourier transform. Shape the target machining Calculating the residual shape by subtracting from Jo, it is characterized by having a step of calculating the said residue difference shape simulated removal.
In the present invention, in the step of calculating the residual shape by the removal simulation, based on a change in unit machining shape that occurs during actual machining that is predicted in advance, for each position on the workpiece surface. It is possible to adopt a configuration including a process in which the unit machining shape is changed to reflect actual machining.
Further, the step of calculating the residual shape includes a process in which the residence time distribution is converted so that the residence time exists along the trajectory through which the polishing tool passes to reflect the actual machining. The configuration can be taken.
Further, in the step of calculating such a residual shape, the shape error component having a short spatial wavelength of the workpiece smoothed by the polishing tool during the actual machining is removed from the target machining shape, and the actual machining is performed. It is possible to adopt a configuration including a process to be reflected.

  According to the present invention, it is possible to obtain a high degree of freedom in calculating a shape close to the actual machining shape, and to satisfy the ability to cope with the target machining shape, the calculation time, and the like.

According to the above configuration of the present invention, it is possible to increase the degree of freedom of calculation while maintaining the advantage of fast Fourier transform.
Conventionally, as described above, (1) the residence time distribution and (3) the residual shape are calculated as described above, as described above, (A) the method using the fast Fourier transform, or (B) the removal simulation. Either one of the methods using the method was used.
On the other hand, in the present invention, (1) the above-mentioned (A) fast Fourier transform is used for calculating the residence time distribution, and (3) the above-mentioned (B) removal simulation is used for calculating the (3) residual shape. The method was adopted. Thereby, it is possible to increase the degree of freedom of the above calculation while maintaining the advantage of the above (A) fast Fourier transform with respect to the capability to calculate the target machining shape and the calculation time.

Embodiments of the present invention will be described below.
FIG. 7 is a diagram for explaining a method using fast Fourier transform for calculating the residence time distribution and using removal simulation for calculating the residual shape in the present embodiment.
In the present embodiment, when calculating the (3) residual shape using the (B) removal simulation, a method of reflecting the actual machining shape is incorporated as follows.
In the present embodiment, a change in unit machining shape that occurs during actual machining is predicted in advance based on experimental data or the like, and the unit machining shape is changed for each position on the workpiece surface.
In the conventional method, due to the restriction for using the fast Fourier transform, the unit machining shape cannot be calculated unless the same machining shape is used in all machining regions. However, in actuality, when the peripheral portion of the workpiece is machined, a part of the polishing tool pops out of the workpiece, the contact between the polishing tool and the workpiece changes, and the unit machining shape changes. Therefore, here, a plurality of unit machining shapes are prepared for each popping amount of the polishing tool, and the unit machining shape is changed in accordance with the popping amount of the polishing tool around the workpiece. This makes it possible to derive a residual shape that is closer to actual machining.

In the present embodiment, the residence time distribution is converted so that the residence time exists along the trajectory through which the polishing tool passes.
In the conventional method, due to the restriction of using the fast Fourier transform, the residence time distribution cannot be calculated unless it is uniformly distributed over the entire processing region. However, since the actual polishing tool continuously moves along a predetermined track, the residence time exists only on this track. Therefore, here, the residence time existing at an arbitrary coordinate is collected into the coordinates on the trajectory through which the nearest polishing tool passes. This makes it possible to derive a residual shape that is closer to actual machining.

In the present embodiment, a filtering process corresponding to the smoothing ability of the processing surface of the polishing tool in the target processing shape is performed.
In actual machining, even if machining is performed with a uniform residence time distribution and constant removal in the entire machining area, components with a shorter spatial wavelength than the polishing tool diameter on the machining surface can be smoothed. It becomes. For example, the amplitude of the spatial frequency λ having the workpiece surface shape before processing is set to 100 [%], and the amplitude of the same spatial frequency λ of the workpiece surface shape after processing is set to X [%]. At that time, when the spatial frequency λ is longer than the diameter of the polishing tool, the amplitude X after processing is about 100% which is substantially the same as that before the processing, but when the spatial frequency λ is shorter than the polishing tool diameter, the amplitude X after processing Becomes a value smaller than 100% before processing. Therefore, on the test piece having the same material shape as the surface to be processed in advance and having various spatial wavelength components, processing is performed with the same polishing tool as that used in actual correction polishing, and the amplitude X [ %] And spatial frequency data.
Based on this data, a filter function for scaling the amplitude for each spatial frequency of the target machining shape is created. By applying a filter corresponding to the smoothing ability of the tool to the target machining shape before subtraction from the calculated machining shape, it is possible to derive a residual shape that is closer to the actual machining.

  The data processing until the residence time distribution is derived is as shown in FIG. 8 in the conventional method (A) using the fast Fourier transform, but in the present embodiment as shown in FIG. To be processed. As a result, in this embodiment, the derivation of the residence time distribution reflecting the actual machining can be performed in a short time with respect to various target machining shapes, and a highly accurate shape correction machining can be performed.

Next, a case where the residence time calculation method according to the present embodiment is used in, for example, a highly accurate free-curved surface polishing apparatus will be described.
These will be described with reference to FIG. 10 showing a flowchart of the operation. First, shape measurement is performed on a workpiece W (hereinafter referred to as a workpiece) (step S1).
As a result, a difference from the design shape, that is, a target machining shape (data D5) is obtained from the obtained workpiece shape data (data D4) and the workpiece design shape (data D3) (step S2).
Next, it is determined whether or not the error shape has reached the target accuracy (step S3).
If it has already reached the target accuracy, it ends. If it has not yet reached, the calculation operation (step S4) for obtaining the residence time from the unit machining shape (data D6) and the target machining shape is performed, and the residence time A distribution (data D8) is obtained.

Next, a calculation operation for obtaining the residence time, which is a feature of the present invention, will be described.
First, a unit machining shape is obtained. In this method, a test piece having the same material shape as that of the surface to be processed is previously polished at a fixed position for a known time with the same polishing tool and polishing conditions as those used in actual correction polishing. And it is obtained by measuring the shape of the polishing recess obtained by this polishing and converting it to per unit time. At this time, for example, if the known time is 600 seconds and the unit time is 1 second, the depth of the polishing recess obtained by polishing at a fixed position of 600 seconds may be multiplied by 1/600. This unit processed shape has a shape as shown in FIG. 11, for example. The center of the unit machining shape is (u, v).

Next, the unit machining shape (data D6) and the target machining shape (data D5) obtained in step S2 are read into the calculation area of the control device 33.
These two data groups are composed of (x, y, z) three-dimensional data per point.
x and y are in the form of equally spaced meshes, z represents the target machining value and the unit machining value, and z values corresponding to the coordinates of (x, y) are respectively set to the target machining value d (x, y), The unit machining value f (x, y) is used. Using these two data groups, a residence time that is a time during which the polishing tool stays on the surface to be processed is calculated.
Similarly, this dwell time is also composed of (x, y, z) three-dimensional data per point, where x and y are equally spaced meshes, z represents the dwell time, and (x, y) The z value corresponding to each coordinate is indicated by the residence time g (x, y).

を設定する。 Next, data processing up to outputting the residence time distribution will be described with reference to FIG.
First, (1) the residence time distribution is calculated. The residence time distribution g _sum (x, y) is calculated from the unit machining shape f (x, y) and the target machining shape d (x, y) in which the polishing tool is in contact with the workpiece throughout the entire area. The unit machining shape f (x, y) is fast Fourier transformed to F (w _x, w _y ). The target machining shape d (x, y) is fast Fourier transformed to D (w _x, w _y ). Filter function from unit machining shape F (w _x, w _y ) using functional

Set.

から滞留時間分布Ｇ（ｗ_x,ｗ_y）を計算する。滞留時間分布Ｇ（ｗ_x,ｗ_y）を逆フーリエ変換し、ｇ（ｘ，ｙ）とする。
Deconvolution integral

The residence time from distribution G (w _x, w _y) is calculated. The residence time distribution G (w _x, w _y ) is inverse Fourier transformed to be g (x, y).

Next, (2) the residence time distribution is corrected.
The residence time distribution g (x, y) calculated in (1) is used as the sum of the residence time distributions calculated up to the previous iteration g _sum (x, y) (initial value of g _sum (x, y)). Is added to 0 in all areas). Thereby, a new residence time distribution g _sum (x, y) is set. In actual processing, since a negative residence time cannot exist, the residence time distribution is converted so that the point value at which the residence time becomes negative becomes 0, and g _sum (x, y) → g _sum '(x, y).

Next, (3) the residual shape is calculated.
The residence time existing on an arbitrary coordinate in the residence time distribution g _sum '(x, y) is aggregated on the coordinates on the trajectory through which the nearest polishing tool passes, and the trajectory through which the polishing tool passes during actual machining is reflected. The residence time distribution is converted into g _sum ' _line (x, y).
The unit machining shape is different from f (x, y) in the case where the previous polishing tool is in contact with the workpiece in the entire region, and the unit machining shape f ₁ (x , Y) to f _n (x, y) are registered.
The target machining shape d (x, y) is subjected to a filter function corresponding to the smoothing ability of the polishing tool obtained by experiment. Thereby, it converts into the target processing shape d '(x, y) reflecting the smoothing capability of the polishing tool.
From the residence time distribution g _sum ' _line (x, y), the unit machining shape f (x, y), and the target machining shape d' (x, y), the residual shape e ( x, y) is obtained.

Next, (4) the residence time distribution is evaluated. It is determined whether the residual shape satisfies the required standard. If not satisfied, (5) reset the residual shape e (x, y) as a new target machining shape d (x, y), prepare for repeated calculation, and (1) stay again Calculate the time distribution.
Subsequently, until the above (4) evaluation of the residence time distribution satisfies the standard for which the residual shape is required, iterative calculation is performed, and (6) residence is obtained by finally calculating g _sum '(x, y). Output time distribution.

Next, an example of a polishing apparatus for carrying out the polishing method in the present embodiment will be described with reference to FIG.
On the bed 10 of the NC polishing apparatus, a Y table 11 is mounted that can reciprocate in the Y direction relative to the bed 10. The movement of the Y table 11 is driven by a Y motor 12. An encoder 13 is attached to the Y motor 12, and the Y direction movement value of the Y table 11 is detected by the encoder 13.
Further, on the Y table 11, an X table 14 that is reciprocally movable in the X direction orthogonal to the Y table 11 is attached. The movement of the X table 14 is driven by an X motor 15. An encoder 16 is attached to the X motor 15, and the X direction movement value of the X table 14 is detected by the encoder 16.
A polishing tank 17 is fixed on the X table 14, and in this polishing tank 17, a support body that rotatably supports a substantially L-shaped holding body 20 that holds the workpiece W in a side view. 18 is fixed.
The drive rotating shaft of the motor 22 attached to the support 18 is coupled to the shaft 19 of the vertical piece 20a of the holding body 20, and the holding body 20 is supported by the motor 22 so as to be rotatable around the axis 19 in the X direction. Has been.
A rotary table 21 that holds the workpiece W is installed on the horizontal piece 20b of the holding body 20, and the rotary table 21 is rotated and detected by a motor and an encoder (not shown).

A column 23 is erected on the X table 14 outside the polishing tank 17, and a guide 24 in the vertical direction (Z direction) is formed on the column 23. A polishing head holder 25 that can reciprocate in the Z direction along the guide 24 is connected to the tip of a rod 32 of an air cylinder 31 as a driving means attached to the upper end of the column 23. The polishing head holder 25 is moved in the Z direction along the guide 24 of the column 23 by driving the air cylinder 31.
Further, the above-described polishing head 26 is attached to the polishing head holding body 25 via an inclination positioning mechanism 27.
A motor 30 for driving the tilt positioning mechanism 27 is attached to the polishing head holding body 25, and the tilt position of the polishing head 26 can be determined via the tilt positioning mechanism 27 by driving the motor 30.

In addition, the polishing tank 17 is filled with a polishing liquid having an appropriate value, and in particular, the stability of the polishing process can be improved by adjusting the particle size of the cerium oxide abrasive grains in the polishing liquid. It is desirable to use a polishing liquid in which a dispersing agent for abrasive grains is agitated.
The control device 33 is for controlling the operation of the polishing device. Information stored in the memory area, the movement values of the Y table 11 and the movement values of the X table 14 from the encoders 13 and 16, etc. A movement value or the like of the rotary table 21 is input. Then, a command is issued to the Y motor 12, the X motor 15, the motor 22, the motor 30, and a drive motor (not shown) of the rotary table 21 that rotates the workpiece W. Alternatively, other than these, a command is issued to the drive motor (not shown) of the polishing head 26, the air cylinder 31, or a motor that replaces the movement thereof. Thereby, each motor can be driven. Specifically, on the surface to be processed according to the polishing scan pattern from the value obtained by the calculation of the dwell time, the polishing scan pattern (data D7) and the design shape (data D3) already stored. A command is issued to scan the polishing tool at a variable speed. By this command, each of the motors described above is driven and the polishing tool is controlled, whereby it is possible to polish the free curved surface with high accuracy.

Examples of the present invention will be described below.
In this example, the polishing method described in the present invention and the embodiment of the present invention was applied, and a workpiece was processed under the following processing conditions.
In this embodiment, the workpiece is a flat surface quartz material having a diameter of 100 mm. Further, a urethane polishing tool having a diameter of 8 mm was used as the polishing tool.
The polishing head rotated the polishing tool, the polishing load was 800 gf, the radius of the rotating motion was 2 mm, and the frequency was 6 Hz.
The abrasive used was cerium oxide, the particle size was 1.25 μm, the medium was generated water, and the concentration was 2 wt%.
The shape correction processing was performed under the above processing conditions.
As a result, the residual shape calculated from the finally calculated residence time distribution was improved, and the actual machining result was also improved.

The figure for demonstrating the basic model of residence time control. The figure for demonstrating the method using the fast Fourier transform in the nonpatent literature 1 which is a prior art example. The figure for demonstrating derivation | leading-out of actual residence time distribution. The figure for demonstrating the residence time distribution calculation by the method using the simulation of the removal in the nonpatent literature 2 which is a prior art example. The figure for demonstrating the residual shape calculation by the method using the simulation of removal. The figure which summarized the pros and cons of the method using the fast Fourier transform of a prior art example, and the method of using a removal simulation. The figure for demonstrating the method of using fast Fourier transformation for derivation | leading-out of residence time distribution in embodiment of this invention, and using the simulation of removal for derivation | leading-out residual shape. The figure explaining the data processing until it derives the residence time distribution by the method using the fast Fourier transform of a prior art example. A figure explaining data processing until deriving residence time distribution by a method of an embodiment of the invention. The flowchart for demonstrating the case where the calculation method of the residence time in embodiment of this invention is applied and used for a grinding | polishing apparatus. The figure which shows the unit processing shape used when demonstrating the calculation method of the residence time in embodiment of this invention. The figure which shows the structure of the grinding | polishing apparatus which enforces the grinding | polishing method in embodiment of this invention.

Explanation of symbols

T: Polishing tool W: Workpiece 10: NC polishing apparatus bed 11: Y table 12: Y motor 13: Encoder 14: X table 15: X motor 16: Encoder 17: Polishing tank 20: Workpiece holder 33: Control device

Claims (4)

Based on the shape of the workpiece, the target machining shape that is the difference between the shape of the workpiece and the design shape, and the unit machining shape that is the machining shape obtained when machining for a unit time,
A residence time for the polishing tool used for processing the workpiece to stay at each coordinate of the workpiece is calculated, and control is performed so that the polishing tool stays at each coordinate of the workpiece during the residence time. A polishing method for polishing a workpiece,
From the unit machining shape and the target machining shape, calculating a residence time distribution necessary for minimizing a residual shape that is an error shape remaining after machining using a fast Fourier transform;
By subtracting the calculated machining shape obtained by superimposing the unit machining shapes for each residence time at each position on the workpiece surface and performing this over the entire machining area from the target machining shape. Calculating the residual shape, calculating the residual shape by removal simulation;
A polishing method characterized by comprising:   In the step of calculating the residual shape by the removal simulation, the unit machining shape is determined for each position on the workpiece surface based on a change in the unit machining shape that is predicted in advance during actual machining. The polishing method according to claim 1, further comprising a process that is changed to reflect actual processing.   In the step of calculating the residual shape by the simulation of the removal, a process in which the residence time distribution is converted to reflect actual machining so that the residence time exists along the trajectory through which the polishing tool passes. The polishing method according to claim 1, further comprising: In the step of calculating the residual shape by the removal simulation, the shape error component having a short spatial wavelength of the workpiece smoothed by the polishing tool during the actual machining is removed from the target machining shape, and the actual machining is performed. The polishing method according to claim 1, further comprising a process that reflects the above.

Images