JP2011036978A - Dressing state determination method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a dressing state determination method which can determine whether or not a grinding stone have achieved a desired straight shape. <P>SOLUTION: The number of cross-points between ultrasonic waves generated in dressing and a plurality of predetermined sample lines is acquired as differential properties in a T method in an MT system. In addition, among the cross-points, the sum of spacings between the cross-points where ultrasonic waves are present at a position above the position of the cross-point is acquired as integral properties in a T method in an MT system. Differential properties and integral properties in predetermined effective sample lines among the plurality of sample lines are standardized by predetermined unit space data. An estimated value of the straightness of the grinding stone achieved by dressing is calculated based on the standardized differential properties and integral properties. When the calculated estimated value satisfies a predetermined threshold value, a desired dressing state is determined to have been achieved. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a method for determining the quality of a dressing state of a grindstone, and more particularly, to a method for determining the quality of a dressing state based on a correlation with a straight shape of a grindstone.

Conventionally, as a method for determining the quality of a grinding wheel dressing state, for example, in Patent Document 1, an ultrasonic wave generated at the time of dressing is detected, and the axis of the outer peripheral surface of the grinding wheel is determined depending on whether or not the intensity of the detection signal is at a substantially constant level. The thing which judged the flatness in a direction is disclosed.
Further, Patent Document 2 detects ultrasonic waves generated during dressing, counts the number of times that the detection signal exceeds and falls below a predetermined threshold during dressing, and determines the dressing state based on the counted value. What has been made is disclosed.

Japanese Patent Laid-Open No. 5-13136 (page 3-7, FIGS. 1 to 12) Japanese Patent Laid-Open No. 11-235665 (page 4-5, FIGS. 1-2)

  However, in any of the determination methods described above, the correlation with the straight shape of the grindstone is not always clear. That is, for example, when performing inner diameter grinding of a nozzle or the like, the grindstone after dressing needs a straight shape at a submicron level, but in the conventional judgment method described above, the quality judgment of dressing is It does not necessarily guarantee a straight shape at the submicron level, and there is a problem that it cannot be reliably determined whether or not a desired straight shape is obtained.

  This invention is made | formed in view of the said actual condition, and provides the dressing state determination method which can determine whether the desired grindstone straight shape was achieved.

In order to achieve the above object of the present invention, a dressing state determination method according to the present invention includes:
A dressing state determination method for determining whether a dressing state is good or bad,
A plurality of sample lines corresponding to the signal waveform are set for the ultrasonic waves generated during dressing, and the number of intersections with the signal waveform in each of the plurality of sample lines is acquired as a differential characteristic of the MT system. In addition, the sum of the intervals between the intersections where the signal waveform is present at a position that exceeds the position of the intersection among the intersections is acquired as an integral characteristic of the MT system, and among the plurality of differential characteristics and integral characteristics, Defined differential characteristics and integral characteristics The corresponding differential characteristics and integral characteristics are used as effective feature items, normalized by unit space data determined in advance, and a grinding wheel by dressing based on the standardized differential characteristics and / or integral characteristics It is configured to calculate an estimated value of straightness and determine that a desired dressing state has been reached when the estimated value satisfies a predetermined threshold value. It become one.

  According to the present invention, unlike conventional ones, it is possible to determine whether the dressing quality is good or not based on whether or not the desired grindstone straightness is satisfied. Therefore, the desired dressing state is efficiently increased without performing unnecessary dressing. There is an effect that it can be secured with reliability.

It is a block diagram which shows the apparatus structural example to which the dressing state determination method in embodiment of this invention is applied. It is the block diagram which showed the functional block of the dressing state determination apparatus in the structural example shown by FIG. It is a flowchart which shows the procedure of the dressing state preparation performed before performing the dressing state determination method in embodiment of this invention. It is a flowchart which shows the procedure of the dressing state determination process performed in the dressing state determination apparatus in embodiment of this invention. It is a wave form diagram which shows an example of the ultrasonic waveform in the case of the dressing obtained by the ultrasonic detection apparatus in embodiment of this invention. FIG. 6 is a waveform diagram showing a main part of an ultrasonic waveform for explaining extraction of differential characteristics and integral characteristics in the ultrasonic waveform shown in FIG. 5. It is explanatory drawing which shows the example which put together the extraction result at the time of extracting a differential characteristic and an integral characteristic in dressing state diagnostic preparation in the table form. It is explanatory drawing which showed the specific numerical example at the time of extracting the differential characteristic and integral characteristic of signal space data in dressing state diagnostic preparation. It is explanatory drawing which shows the example which put together the result of normalizing the differential characteristic and integral characteristic of the signal space data extracted in the dressing state diagnostic preparation in a tabular form. It is explanatory drawing which shows the result of normalization of the differential characteristic and integral characteristic of the signal space data shown by FIG. It is explanatory drawing which showed the example of the proportionality constant (beta) and SN ratio (eta) with respect to the normalized value of the signal space data shown by FIG. It is explanatory drawing which shows the example at the time of adding the estimated value of each member with respect to the example shown by FIG. It is explanatory drawing explaining the example which allocated all 26 items to the 2 level system orthogonal table of L32. It is the bar graph which arranged and represented 2 level system SN ratio (eta) average difference in order of the item with the largest value.

Hereinafter, embodiments of the present invention will be described with reference to FIGS.
The members and arrangements described below do not limit the present invention and can be variously modified within the scope of the gist of the present invention.
First, a configuration example of the dressing state determination apparatus in the embodiment of the present invention will be described with reference to FIG.
First, the grinding apparatus 201 itself shown in FIG. 1 is a known and well-known one, and an example having a dressing function is shown in FIG. This figure mainly shows the components for dressing the grindstone 1 with the dresser head 211, and the illustration of the configuration of the grinding portion of the workpiece by the grindstone 1 is omitted.

The grinding device 201 has a configuration in which an ultrasonic detection device 301 that detects ultrasonic waves generated during dressing is added.
That is, the ultrasonic detection apparatus 301 is roughly divided into an ultrasonic sensor 311, a transmission unit (not shown), and a reception unit 312. In the embodiment of the present invention, the ultrasonic sensor 311 is illustrated. It is arranged in the dresser head 211 together with the transmitter not to be performed.

The transmission unit (not shown) and the reception unit 312 in this example are configured to wirelessly transmit and receive the detection signal of the ultrasonic sensor 311, and the detection signal of the ultrasonic sensor 311 demodulated by the reception unit 312. Is input to a dressing state determination device 101 (denoted as “JUD-UNIT” in FIG. 1) 101.
The determination result of the dressing state by the dressing state determination device 101 is supplied to the operation control of the grinding device 201 by the grinding control unit 202 (denoted as “CONT” in FIG. 1).
The ultrasonic detection apparatus 301 in FIG. 1 is configured to use radio for transmission / reception of signals between a transmission unit (not shown) and the reception unit 312. Of course, the ultrasonic detection apparatus 301 needs to be limited to such a configuration. There may be a wired type.

  The dressing state determination device 101 includes, for example, a microcomputer (not shown) having a known and well-known configuration, a storage element (not shown) such as a RAM and a ROM, and a grinding device 201. An interface circuit (not shown) or the like for interfacing is mainly configured as a component.

FIG. 2 is a functional block diagram showing functions configured by software for executing the dressing state determination process in the dressing state determination apparatus 101, which will be described below.
First, an AE wave signal (determination AE wave signal) referred to when dressing a grindstone to be diagnosed is input to the dressing state determination device 101 via the ultrasonic detection device 301 and is input. A predetermined feature amount is extracted from the determination AE wave signal by the feature amount extraction means (details will be described later).

  Next, based on the feature amount extracted by the feature amount extraction unit, various calculations for determining the grindstone straightness are calculated by the grindstone straight shape estimated value calculation unit (details will be described later), and desired by the straightness determination unit. It is determined whether or not the grindstone straightness has been reached. The determination result is used for controlling the dressing operation in the grinding apparatus 201.

In FIG. 3, the procedure of the dressing diagnosis preparation process, which is a preliminary preparation process necessary for executing the dressing state diagnosis process in the dressing state determination apparatus 101, is executed in the dressing state determination apparatus 101 in FIG. The processing procedure of the dressing state diagnosis process is shown respectively. Hereinafter, these processing procedures will be described with reference to FIG.
First, the dressing state diagnosis method in the embodiment of the present invention will be generally described.
In the dressing state diagnosis method according to the embodiment of the present invention, whether or not the straight shape of the grindstone by dressing satisfies a desired straightness is determined using the T method which is one method of the MT system. It is a thing.

  In such a dressing state diagnosis method, an ultrasonic signal obtained by dressing is analyzed based on the T method, a predetermined specific feature amount is selected from the analysis data, and the extracted feature amount is selected. The quality of the dressing state is determined by whether or not the estimated value of the straight shape of the grindstone calculated based on the above satisfies a threshold value corresponding to the desired straightness.

In order to carry out such a dressing state diagnosis method, dressing diagnosis preparation processing is required in order to preset items for data selection criteria at the time of diagnosis. FIG. 3 shows the procedure. Hereinafter, the dressing diagnosis preparation process will be described with reference to FIG.
First, dressing is performed on a grindstone previously selected as a grindstone for setting a unit space in the T method, and an ultrasonic (AE) waveform generated at that time is measured and acquired (see step S100 in FIG. 3).

Here, the AE waveform is not so-called raw data, but is a waveform that has been amplified and filtered in the receiving unit 312, further subjected to detection processing, and then subjected to envelope processing.
Also, the AE waveform in the dressing diagnosis preparation process is not input to the grinding control unit 202, but is input from the receiving unit 312 to a well-known and well-known waveform measuring device (not shown), unlike the case of the dressing diagnosis process described later. Shall be
Further, as the grindstone used for dressing in this dressing diagnosis preparation process, a standard one is used among those having the same roughness and hardness as the grindstone to be diagnosed in the dressing diagnosis process described later.

Next, the straight shape of the grindstone after dressing is measured (see step S150 in FIG. 3).
In the embodiment of the present invention, the straight shape of the grindstone is measured by transferring the grindstone after dressing to an appropriate workpiece and measuring the transferred shape.
Specifically, the grindstone after dressing is subjected to normal oscillation grinding, and, for example, inner diameter grinding is performed to a position where the workpiece diameter is 4 mm. At this time, the cutting is performed at a cutting speed as slow as possible or with a manual pulse so as not to change the shape of the grindstone surface. After cutting to a position where the workpiece diameter is 4 mm, the cutting is once returned from the workpiece, then the grindstone is moved to the center position of the workpiece, and the workpiece is machined by several microns by plunge cutting (manual pulse). As a result, a step is formed only on the grindstone contact portion on the workpiece machining surface, and this shape is measured as the grindstone straightness.

Next, a unit space (in other words, a reference space) is set by using the grindstone straightness measured as described above as unit space data (see step S200 in FIG. 3).
Next, the signal space is set (see step S250 in FIG. 3).
That is, dressing is performed for a plurality of grindstones selected in advance for signal space setting, and the AE waveform and grindstone straightness are measured, and this is used as signal space data.

  In the embodiment of the present invention, for two types of grindstones having the same roughness and hardness as the grindstone used for setting the unit space described above and having different compositions, dressing is performed using three grindstones for the signal space. went. The dressing conditions are the same as the dressing conditions for the grindstone for setting the unit space (see step S100 in FIG. 3).

Next, feature items are extracted from the data in the unit space and the signal space (see step S300 in FIG. 3).
The feature item is data that should be referred to as basic data for calculating an estimated value by the T method as described later.
First, AE waveform data is filtered by a moving average (for example, 10 times), and feature items are extracted from the filtered AE waveform data as described below.

FIG. 5 shows an example of the AE waveform data after the above-described filtering, and the feature item extraction procedure will be described below with reference to FIG.
First, for the AE waveform data, the position of the average output value is determined as the reference position. In FIG. 5, the one-dot chain line with the symbol Lc represents the reference position in this AE waveform data example, and in this example, the reference position is the position of the AE output 2.0V. .

With respect to this reference position Lc, a plurality of lines (hereinafter referred to as “sample lines” for convenience) are drawn parallel to the time axis within a predetermined range (see FIG. 5), and on each sample line, The intersection of the sample line and the AE waveform is extracted as a feature item as described below (see FIG. 5). In FIG. 5, the sample line is indicated by a two-dot chain line.
The interval between the plurality of sample lines is increased as the distance from the reference position increases. In other words, the determination accuracy is improved by setting the interval between the sample lines to be smaller as the reference position is approached.
In the embodiment of the present invention, the predetermined range above and below the reference position Lc is ± 0.3V, and the interval between the sample lines is 0.05V.
This is only an example, and the predetermined range above and below the reference position Lc and the interval between the sample lines are suitable for use as data in consideration of the level of the detected AE wave, the output level of the measuring instrument, and the like. It should be determined appropriately so as to be within the range.

In the embodiment of the present invention, two types of feature items are the number of intersections between the sample line and the AE waveform and the sum of the lengths of the intersections between the AE waveform on the sample line.
The number of intersections between the sample line and the AE waveform is referred to as “differential characteristics”, and the sum of the distances between the intersections with the AE waveform on the sample line is referred to as “integration characteristics”. This is a characteristic item often used in the MT system. The “distance” in the integral characteristic may be either the length of the line segment or the time between them, but in the embodiment of the present invention, it will be described as the length of the line segment.

FIG. 6 shows an AE waveform diagram for more specifically explaining the extraction of the differential characteristic and the integral characteristic at the reference position in the AE waveform shown in FIG. 5. The contents will be described.
First, in FIG. 6, for example, black circled points with reference numerals X21 and X22 indicate intersections between the AE wave and the sample line at the reference position. In FIG. 6, black circles are omitted from some intersections from the viewpoint of simplifying the drawing and facilitating understanding.
The differential characteristic is defined as the number of intersections on each sample line.

On the other hand, the integral characteristic is defined as the sum of the lengths between the intersections having the AE wave output among the intersections as the differential characteristics, that is, the lengths between the intersections where the AE wave exists above the sample line. Is.
That is, in FIG. 6, for example, the point between the black circle point with the symbol X21 and the black circle point with the symbol X22 is between the intersections where the AE wave exists, and the length between the two intersections, That is, the length or time of the actual line segment on the graph is the length or time between the intersections.

On the other hand, in FIG. 6, there is an AE wave between the black circle point with the symbol X22 and the black circle point with the symbol X23 because the level of the AE wave is lower than the sample line at this position. The length between the two intersections is not treated as an integral characteristic.
The integral characteristic is defined as the sum of the lengths of the plurality of line segments thus obtained on one sample line (hereinafter referred to as “line segment sum” for convenience).
The above-described extraction of the differential characteristic and the integral characteristic is performed for each of the unit space data and the signal space data, and these extraction results are summarized in a table having a format as shown in FIG. 7, for example.

FIG. 7 is a generalized representation of the structure of the table. To explain each content, first, “member” means an individual sample, that is, a grindstone, and each calculation in the row of “member”. The number is an identification number assigned to each grindstone in order to classify each grindstone.
The number of samples (grinding stones) used for setting the unit space, that is, the number of members n is n ≧ 1, and in the embodiment of the present invention, n = 1.
On the other hand, the number of members 1 in the signal space is preferably plural, and in the embodiment of the present invention, the number of members 1 is 6.

In FIG. 7, “output” means a physical quantity to be diagnosed using this T method, that is, the straightness of the grindstone. In the “output” row, y1 to yn are members of the unit space. The straightness for each member, and y′1 to y′l are the straightness for each member of the signal space, and are actually measured values.
The “average value” in the unit space column is the average value of the outputs of each member of the unit space described above.

In FIG. 7, items 1 to k are items representing the differential characteristics and integral characteristics for each sample line described above with reference to FIG.
That is, item 1 to item (k / 2) are items representing the number of intersections with the AE wave in each sample line from the higher side of the sample line, and item (k / 2 + 1) to item k are sample lines. This is an item representing the sum of line segments in each sample line from the higher position of.

In FIG. 7, for example, x11 to xn1 in the row of the item 1 in the unit space are the values of the feature extraction items of each member, that is, the values of the differential characteristics described above in the embodiment of the present invention, In other words, the number of intersections with the AE wave on the corresponding sample line. The same applies to the values of x in the other item rows.
This is basically the same in the signal space. That is, in FIG. 7, for example, x′11 to x′n1 in the row of item 1 is the number of intersections with the AE wave for each member on the sample line at the highest position.

FIG. 8 shows a specific example of the signal space data based on the format of FIG. 7, and will be described below.
The example shown in FIG. 8 is an example of the case where the number of members in the signal space is l = 6 and feature item extraction is performed using 13 sample lines.
The grindstone straight shape is an actual measurement value of each member as described in FIG. 7, and the unit is “μm”.

In FIG. 8, items 1 to 13 are differential characteristics, and items 14 to 26 are integral characteristics. In the example of FIG. 8, items 6 to 25 are omitted.
The integral characteristic in this example is the sum of line segments as described above with reference to FIG. 7, and the unit is “mm”.
In the case of unit space, although the specific numerical value is different, it is obtained as shown in FIG. 8, but in the embodiment of the present invention, the number of members is 1 as described above. is there.

Here, returning to the description of FIG. 3, after extracting the feature items (see step S300 in FIG. 3) as described above, the signal data is normalized (see step S350 in FIG. 3). ).
Normalization of signal data is performed using unit space data.
FIG. 9 illustrates an example in which the result of normalization is generalized and expressed in a tabular form. Hereinafter, the normalization procedure will be described with reference to FIG.

Normalization is also performed on the output, and in FIG. 9, M1 to Ml are normalized output values of each member.
This output normalization is obtained by the following equation (1).

  Mj = y'j-M0 (j = 1, 2,... L) Equation 1

  Here, y′j (j = 1, 2,..., L) is an output of each member of the signal space before normalization (see FIG. 7). M0 is an average value of the output of the unit space (see FIG. 7).

  Next, in FIG. 9, the values X11 to Xlk of each item of each member are normalized item values obtained by the following equation 2.

  Xij = x'ij-xj Equation 2 (where i = 1, 2,... L, j = 1, 2,... K)

Here, x′ij is each data (each item value) in the signal space before normalization, and xj is an average value of each item value in the unit space before normalization (see FIG. 7). .
That is, the normalized item value is obtained as a difference between the signal space data and the unit space data.
When the specific example of the signal space data shown in FIG. 8 is normalized by such normalization procedure, the result is as shown in FIG.

After normalization as described above, the proportionality constant β and the square ratio are calculated for each item value in the standardized signal space (see step S400 in FIG. 3).
A specific calculation method will be described below with reference to FIG.
FIG. 11 shows an example in which the proportionality constant β and the calculation result of the square ratio are expressed in general terms. The proportionality constant β and the calculation formula for the square ratio will be described with reference to FIG. .
For example, the proportionality constant β1 is a proportionality constant for the item 1, which is obtained by dividing the sum of the products of the output values M1 to Ml of each member and the value of the item 1 of each member by the effective divisor. Desired. That is, it is obtained by the following formula 3.

  β1 = (M1 · X11 + M2 · X21 +... + Ml · Xl1) / r Equation 3

Here, M1 to Ml are normalized output values of each member, and X11 to Xl1 are normalized values of item 1 of each member (see FIG. 9).
The effective divisor r is defined as the sum of the squares of the outputs M1 to Ml of each member, and is specifically obtained by the following equation 4.

r = M1 ² + M2 ² +... + Ml ^2.

Other proportionality constants β2 to βk are obtained in the same manner according to the above-described calculation procedure of β1.
On the other hand, with respect to the square ratio, for example, taking the square ratio η1 as an example, the square ratio is obtained by the following equation (5).

η1 = (1 / r) (S _β1 −Ve1) / Ve1 Equation 5

However, Equation 5 is valid only when S _β1 > Ve1 is satisfied, and is defined as η1 = 0 when S _β1 ≦ Ve1.
Here, S _β1 is referred to as “variation of the proportional term”, and is obtained by the following Equation 6.

S _β1 = (M1 · X11 + M2 · X21 +... + Ml · Xl1) ² / r.

That is, the variation S _β1 of the item 1 is obtained by dividing the result of squaring the sum of the products of the output values M1 to Ml of each member and the value of the item 1 of each member by the effective divisor r.
Further, Ve1 is called “error variance”, and is obtained by the following equation (7).

  Ve1 = Se1 / (l-1) Equation 7

In Equation 7, “l” is the number of members in the signal space (see FIG. 11).
In Expression 7, Se1 of the first term is called “error fluctuation”, and is obtained by Expression 8 below.

Se1 = ST1-S _β1 Formula 8

In Equation 8, S _{β1 in} the second term is the variation of the proportional term obtained by Equation 6 above. Further, ST1 of the first term of Equation 8 is called “total variation” and is obtained by Equation 9 below.

ST1 = (X11 ² + X21 ² +... + Xl1 ² )...

  That is, ST1 is obtained as the sum of the squares of the normalized values (see FIG. 9) of item 1 of each member. When η2 is obtained, the total variation is ST2, and is obtained as the sum of the squares of the normalized values (see FIG. 9) of item 2 of each member. Hereinafter, the total variation ST when η3 to ηk are respectively obtained is obtained in accordance with this.

After obtaining the proportionality constant β and the square ratio η as described above, an estimated value for each member's output, that is, an estimated value of straightness is calculated based on the proportionality constant β and the square ratio η (see step S450 in FIG. 3).
FIG. 12 shows an example in which an estimated value is added to FIG.
In general, the estimated value of M is given a symbol “^” above M, but in this specification, for convenience of description, “M ^” is used for convenience.
Regarding the calculation of the estimated value, for example, the estimated value Mi ^ of the member i (i = 1, 2,..., L) is calculated by the following equation 10.

  Mi ^ = {η1 × (Xi1 / β1) + η2 × (Xi2 / β2) + ... + ηk × (Xik / βk)} / (η1 + η2 +++ ηk) Equation 10

That is, the estimated value Mi ^ is obtained as the product of the normalized value of each item and the square ratio η divided by a proportionality constant and the division result divided by the sum of the square ratio η of each item. .
For example, in the example shown in FIG. 12, when i = 1, that is, the estimated value M1 ^ of member 1 is obtained as follows.

  M1 ^ = {η1 × (X11 / β1) + η2 × (X12 / β2) + ... + ηk × (X1k / βk)} / (η1 + η2 +++ ηk)

  Next, a correlation signal-to-noise ratio η representing the correlation between the estimated value calculated as described above and the actual measurement value is calculated by the following equation 11 (see step S500 in FIG. 3).

η (db) = 10 log {(1 / r) (S _β −Ve) / Ve} Expression 11

Here, r is referred to as an “effective divisor” as already described, and is obtained by Equation 4. S _β is referred to as “variation of proportional term” and is obtained by the following equation (12).
Thus, it can be said that the correlation SN ratio η is obtained by dividing the difference between the error fluctuation and the error variance by the effective divisor, and further dividing the division result by the error variance and converting it to a logarithm.

S _β = L ² / r Equation 12

  Here, L is obtained by the linear form of Equation 13 below.

  L = M1, M1 ^ + M2, M2 ^ + ... + Ml, Ml ^ ... Equation 13

That is, L is the sum of products of the output value of each member and its estimated value.
Further, in Equation 11, the error variance Ve is obtained by Equation 14 below.

  Ve = Se / (l-1) Equation 14

  Here, Se is referred to as “error variation”, and is obtained by the following Expression 15.

Se = ST−S _β Formula 15

 Here, ST is referred to as “total variation”, and is obtained by the following equation (16).

ST = (M1 ^ ² + M2 ^ ² + ... + Ml ^ ² ) ... Equation 16

In Equation 15, S _β is the “variation of the proportional term” obtained by Equation 12 above.
Thus, the error fluctuation Se is obtained as the difference between the total fluctuation and the fluctuation of the proportional term. The total variation is obtained as the sum of the squares of the estimated values of the members.
The correlation SN ratio η obtained in this way means that the larger the value, the higher the correlation between the estimated value and the actually measured value, that is, the higher the estimation accuracy.

After obtaining the correlation signal-to-noise ratio η representing the correlation between the estimated value and the actually measured value as described above, it is determined as follows which combination of items is optimal when estimating the straightness. (See step S550 in FIG. 3).
That is, in the embodiment of the present invention, first, an evaluation is made as to which items are highly effective for estimation using a two-level orthogonal table. Here, “item” means a differential characteristic or an integral characteristic for each sample line, and specifically corresponds to items 1 to k in FIG. 12, for example.
Specifically, for example, when the number of items is 26 as illustrated in FIG. 8, it is preferable to allocate all 26 items to the L32 two-level orthogonal table. FIG. 13 shows an example in which all 26 items are assigned to the L32 two-level orthogonal table, and this example will be described with reference to FIG.

First, in FIG. 13, the notation “1” indicates that the item is the first level, that is, the item is used, and the notation “2” indicates that the item is the second level, That is, the item is not used.
After performing the allocation as shown in FIG. 13, the SN ratio based on the previous equation 11 is calculated for each row, that is, for each of number 1 to number 32 (for each factor). In FIG. 13, the SN ratio calculated in this way is described as “total estimation SN ratio”.
For example, in the example shown in FIG. 13, since the number 1 row has all 26 items as the first level, the SN ratio of this row matches the value calculated in the previous step S500. It becomes.

  In FIG. 13, for example, in the row number 2, eight items X1 to X4, X6, X7, X10, and X16 are the first level. Therefore, in calculating the SN ratio η based on Expression 11, only the 8 items are used. Specifically, for the calculation of the estimated value Mi ^ of each member required when calculating the SN ratio η using Equation 11 (see Equation 10), only the 8 items are used. Become.

Next, for the SN ratio η calculated as described above, for each item, a difference between the two is calculated as a comparison of the average of the first level SN ratio η and the average of the second level SN ratio η. Determine the effectiveness of.
That is, in the example shown in FIG. 13, for example, in the case of item X1, the average of the SN ratios η at the first level is obtained by adding the SN ratios η from number 1 to number 16 and calculating the SN ratio η. It is obtained as a result of dividing by 16 which is the number of η. In addition, the average of the SN ratios η at the second level is obtained as a result of adding the SN ratios η from No. 17 to No. 32 and dividing the result by 16 which is the number of the SN ratios η. As a result, in item X1, (average SN ratio η at the first level) − (average SN ratio η at the second level) = 0.08 db.

For example, in the case of item X2, the average of the SN ratio η at the first level is obtained by adding the SN ratios η from number 1 to number 8 and the SN ratios η from number 17 to number 24, Is divided by 16 which is the number of these SN ratios η. On the other hand, the average of the SN ratios η in the second level is obtained by adding the SN ratios η from No. 9 to No. 16 and the SN ratios η from No. 25 to No. 32, and adding them. It is obtained as a result of dividing by 16 which is a number.
In the same manner for other items, the difference between the average of the SN ratio η at the first level and the average of the SN ratio η at the second level in each item is obtained.

Next, the difference between the average of the SN ratio η at the first level and the average of the SN ratio η at the second level in each item calculated as described above (hereinafter referred to as “two-level system SN ratio η average for convenience). Items are arranged in descending order of difference. In FIG. 14, the two-level SN ratio η average difference is obtained for each item as described above for the example of FIG. 13, and based on the result, the items are in descending order of difference, in other words, in descending order of effectiveness. In addition, the example in which each two-level S / N ratio η average difference is represented in a bar graph is shown.
In the example of the figure, items X19 and X24 have the highest gain (2 level SN ratio η average difference).
On the other hand, the lower 11 items are all negative values, and it is assumed that these are ineffective in estimating the straightness or deteriorate the estimation accuracy.

Next, the SN ratio η based on the previous equation 11 is calculated for some combinations of the higher level items of the two-level SN ratio η average difference, and whether or not the calculated value exceeds a predetermined target reference value. Based on the determination, an effective feature item is selected (see step S600 in FIG. 3).
That is, in the examples shown in FIGS. 13 and 14, (1) the top 15 items of the two-level system SN ratio η (in other words, all items for which the average difference of the two-level system SN ratio η is positive) , (2) the upper 7 items, (3) the upper 4 items, and (4) the SN ratio η based on Equation 11 when only the respective items are used is calculated again.
In the case of the example of FIG. 14, the top two items (X19, X24) are the same value and are first in the same column, so the top four items mean that the top three items are taken. Further, how to combine the upper items is not limited to the above example, and should be appropriately selected.

Then, it is determined whether or not the SN ratio η in each item combination calculated as described above exceeds a predetermined target reference value, and the item used for calculating the SN ratio η exceeding the predetermined target reference value Is selected as an effective feature item.
Here, the predetermined target reference value should be appropriately set according to the target straightness.
The item defined as the effective feature item is an item extracted in the dressing diagnosis process described below, and before the execution of the dressing diagnosis process described below, by a method predetermined in the dressing state determination device 101, It is necessary to write (specify) an item number to be an effective feature item.
This completes the dressing diagnosis preparation.

Next, the dressing diagnosis process will be described with reference to FIG.
First, in the previous dressing diagnosis preparation, the receiving unit 312 and the dressing state determination apparatus 101 are in an off-line state, that is, the output of the receiving unit 312 is input to a separate measuring device. In the execution of the process, it is assumed that the output of the reception unit 312 is input to the dressing state determination apparatus 101. As a result, the dressing state determination device 101 determines the quality of the dressing in real time as will be described later, and the determination result is input to the grinding control unit 202 of the grinding device 201 and used for dressing operation control. .

  Under such a premise, first, the grinding device 201 starts dressing for the grindstone to be diagnosed. The ultrasonic wave generated at the start of dressing is detected by the ultrasonic sensor 311, wirelessly transmitted by a transmission unit (not shown), and received and demodulated by the reception unit 312. Then, the ultrasonic signal demodulated by the receiving unit 312 is input to the dressing state determination device 101 which is similarly in an operation start state when the grinding device 201 starts operating, and the dressing state determination processing program in the dressing state determination device 101 The determination AE wave signal is read by starting (see step S802 in FIG. 4).

In the dressing state determination apparatus 101, the reading of the AE wave signal is performed, and at the same time, the differential characteristic and the integral characteristic are extracted from the read AE wave signal as described above with reference to FIGS. For example, according to the format as shown in FIG. 8, it is stored in a predetermined storage area of the dressing state determination apparatus 101.
In the extraction of the differential characteristic and the integral characteristic, desired values are set in advance for the reference position Lc and the interval between the sample lines.

Next, feature amount extraction is performed in the dressing state determination apparatus 101 (see step S804 in FIG. 4).
That is, the effective feature item specified in advance in the dressing diagnosis preparation is extracted from the differential characteristic and the integral characteristic obtained as described above. For example, when item 1 and item 5 (see FIG. 8) are selected as effective feature items in the previous dressing diagnosis preparation (see step S600 in FIG. 3), that is, in other words, the sample line corresponding to item 1 When the sample line corresponding to item 5 is selected as an effective sample line, in step S804, data corresponding to item 1 and item 5 is extracted.

Next, normalization is performed on the data of the items extracted as described above (see step S806 in FIG. 4).
In other words, the extracted item data is normalized using the unit space data, similarly to the normalization of the signal data based on the unit space data previously performed in the dressing diagnosis preparation. The unit space data in this case is set at the time of dressing diagnosis preparation.
Specifically, in the dressing state determination apparatus 101, the determination signal data is normalized based on Equation 2 above. In this case, x′ij in Expression 2 is each data obtained in step S804. Further, xj in Equation 2 is an average value of each item value in the unit space before normalization used in dressing diagnosis preparation (see FIG. 7).

Next, an estimated value for the determination signal data is calculated (see step S808 in FIG. 4).
That is, the estimated value is calculated based on the previous equation 10 in the dressing state determination apparatus 101. In this case, the proportionality constant β and the SN ratio η are values obtained in the dressing diagnosis preparation (see FIG. 11), and Xij (i = 1, 2,..., J = 1, 2,. k) is a normalized value of each item calculated in step S806.

Next, it is determined whether the dressing state is acceptable (see step S810 in FIG. 4).
That is, in the dressing state determination apparatus 101, it is determined whether or not the estimated value obtained as described above satisfies a predetermined threshold value, and when it is determined that the threshold value is satisfied (in the case of YES) In this case, it is determined that the dressing state is acceptable, and a signal for permitting the dressing to be finished is output to the grinding control unit 202, and the series of processing ends. Here, it is preferable that the threshold value is set to a certain allowable range with respect to the desired grindstone straightness.

  On the other hand, when the dressing state determination apparatus 101 determines that the estimated value does not satisfy a predetermined threshold value (in the case of NO), a signal indicating that the dressing is defective is output to the grinding control unit 202. A series of processes will be terminated. In this case, as a signal output from the dressing state determination apparatus 101 to the grinding control unit 202, not only a signal indicating a dressing failure but also a difference between the estimated value and the threshold value is output, and the grinding control unit 202 continues the dressing. It is preferable to be able to determine whether or not it is necessary.

  In the dressing state diagnosis process according to the embodiment of the present invention described with reference to FIG. 4, the feature amount extraction means in FIG. 2 is performed in steps S802 and S804, and the grindstone straightness in FIG. 2 is displayed in steps S806 and S808. The shape estimation calculation means is the straightness determination means in FIG. 2 realized by steps S810 to S814.

DESCRIPTION OF SYMBOLS 101 ... Dressing state determination apparatus 201 ... Grinding apparatus 202 ... Grinding control part 301 ... Ultrasonic detection apparatus 311 ... Ultrasonic sensor 312 ... Reception part

Claims (3)

A dressing state determination method for determining whether a dressing state is good or bad,
A plurality of sample lines corresponding to the signal waveform are set for the ultrasonic waves generated during dressing, and the number of intersections with the signal waveform in each of the plurality of sample lines is acquired as a differential characteristic of the MT system. In addition, the sum of the intervals between the intersections where the signal waveform is present at a position that exceeds the position of the intersection among the intersections is acquired as an integral characteristic of the MT system, and among the plurality of differential characteristics and integral characteristics, Defined differential characteristics and integral characteristics The corresponding differential characteristics and integral characteristics are used as effective feature items, normalized by unit space data determined in advance, and a grinding wheel by dressing based on the standardized differential characteristics and / or integral characteristics Calculating an estimated value of straightness, and determining that a desired dressing state has been reached when the estimated value satisfies a predetermined threshold; Dressing state determining how. Effective feature items are dressing the grindstone for the unit space of the MT system and the grindstone for the signal space, respectively, acquiring ultrasonic waves generated in each, and for each signal waveform, A plurality of sample lines corresponding to the signal waveform are set, the number of intersections with the signal waveform in each of the plurality of sample lines is acquired as a differential characteristic of the MT system, and the position of the intersection among the intersections While obtaining the sum of the intervals between the intersections where the signal waveform is present at a position above the integral characteristic of the MT system, and using the differential characteristic and integral characteristic obtained for the grindstone for the unit space as unit space data, The differential characteristics and integral characteristics acquired for the signal space grinding wheel are used as signal data, and the unit space data is based on the signal data. Performing normalization based on the T method of the MT system, calculating a proportionality constant β and a square ratio based on the T method for the normalized data of each differential characteristic and integral characteristic, and calculating the proportionality constant β, the square ratio, And, based on the normalized data, an estimated value of straightness based on the T method for the grindstone defined for the signal space is calculated, and then the correlation between the calculated estimated value and the actually measured value While calculating the correlation SN ratio η representing the relationship,
The plurality of sample lines are numbered in ascending order from 1 in order to classify the plurality of differential characteristics and integral characteristics constituting the signal data corresponding to the plurality of sample lines set for acquiring the characteristics. The differential characteristic or the integral characteristic acquired in each sample line according to the numbering is divided in order from item 1, and each item is set as a level in the two-level orthogonal table, while being defined for the signal space. Each whetstone is assigned to the two-level orthogonal table as a factor in the two-level orthogonal table, and the correlation SN ratio η for each of the factors is obtained. Thereafter, the correlation SN ratio η of the first level is determined for each item. A difference between the average value and the average value of the correlation SN ratio η of the second level is obtained, a plurality of combinations are arbitrarily set for a plurality of upper items in the calculation result, and a correlation SN ratio η in each item combination is set. And its calculated The differential characteristic and the integral characteristic corresponding to the item used for calculation of the correlation SN ratio η in which the correlation SN ratio η exceeds a predetermined target reference value are selected as effective feature items. The dressing state determination method according to 1.   The sample lines are set at a plurality of positions above and below the average value with the position of the average value of the output level of the ultrasonic wave as a reference position, and as the distance from the average value increases, the interval between the sample lines is coarse. The dressing state determination method according to claim 2, wherein the dressing state determination method is set to be.

Images