Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 26 (2015) 718 723

12th Global Conference on Sustainable Manufacturing

Evaluating and Influencing Dressing Results by Changing the Grain Size

Distribution Based on Statistical and Experimental Investigations
A. Fritsche*, F. Bleicher*
*
Institute for Manufacturing Technique and High Power Laser Technology
Management, Vienna University of Technology, Austria

* Corresponding author. Tel.: +43-1-58801-311384; fax: +43-1-58801-31195. E-mail address: fritsche@ift.at

Abstract

Continuous improvements in product design and production processes are crucial for manufacturing companies in order to remain sustainable
and competitive. In this respect the grinding industry is seeking improvements in more durable dressing results of grinding wheels as one major
governing factor. Tackling this obstacle can be achieved via changing the mixing ratio between different grain sizes and pore volumes. Thus,
the reduction of dressing operations leads to higher productivity and lower material consumption. The determination of the grinding wheel
topography after dressing and grinding is analyzed by an optical 3D-micro-coordinate measurement system. A statistical approach is presented
to estimate the real cutting area of grinding wheels after dressing, according to wheel composition and specific deep grinding working
parameters. With respect to experimental results new compositions for grinding wheels are proposed.

2014
2015 The
The Authors.
Authors. Published
Published by
by Elsevier
Elsevier B.V.
B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of Assembly Technology and Factory Management/Technische Universitt Berlin.
Peer-review under responsibility of Assembly Technology and Factory Management/Technische Universitt Berlin.
Keywords: Grain distribution, Wheel deflection, Real contact length

1. Introduction Geometrical boundary conditions 2. Geometrical boundary conditions

Dressing of silicon carbide and aluminum oxide abrasives is 2.1. Testing parameters
one of the most challenging tasks in grinding technology [1].
This pre-manufacturing operation is the decisive factor for The experiments and calculations were conducted with an
wear resistance and surface quality. Thus, the stochastic nature aluminum oxide wheel of 126 mm diameter and 20 mm
of grain distribution must be taken into consideration in the width. Moreover, the exact wheel composition is listed in
research of grinding processes [2]. Predicting the number of Table 1. The respective abrasive mesh and sieve openings are
cutting grains and, in this respect, the emerging cutting area, provided in Table 2.
are crucial factors for assessing the quality of dressing and
conditioning operations. In addition, the previously mentioned Table 1: Percentage of bond, grain and pores in abrasive
cutting area is of particular interest because this number
indicates the effective available cutting area depending on the Mixing Wheel
Grain Size Bond Grain Pores
chosen working parameters [3]. Understanding the interaction ratio specifi-
[mm] [vol.%] [vol.%] [vol.%]
between grain sizes, pore and bond volume makes it possible [wt.%] cation
Al-
to design a wheel with more predictable behavior [5-6]. F60/70 50/50 Oxide 9 40 51
wheel

2212-8271 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of Assembly Technology and Factory Management/Technische Universitt Berlin.
doi:10.1016/j.procir.2014.07.078
 A. Fritsche and F. Bleicher / Procedia CIRP 26 (2015) 718 723 719

Table 2: Abrasive mesh dmax, dmin and dM for standard wheels according to FEPA

Grain size [mm] F20 F24 F30 F36 F46 F54 F60 F70 F80 F90 F100
dmax 1,18 0,85 0,71 0,60 0,425 0,355 0,30 0,25 0,212 0,18 0,15
d min 0,85 0,60 0,50 0,425 0,30 0,25 0,212 0,18 0,15 0,125 0,106
dM 1,0 0,75 0,63 0,53 0,39 0,32 0,27 0,22 0,18 0,155 0,125

2.2. Geometrical contact length 3. Statistical calculations

The geometrical and the real contact lengths are the governing
attributes of a grinding process [4]. Therefore Figure 1 shows
the interaction between the cutting speed (vw=25 m/s), the
infeed (a0=0,5 mm) and the forward feed (vt=1800 mm/min).
By means of these parameters it is possible to describe the
geometrical behavior of a grinding wheel during the grinding
process.
3.1. Determine the grain specific parameters
The mean grain diameter dM 60\70 is one of the most significant
parameters, which takes into account that two different grain
sizes at a 50/50 mixing ratio (see Table 1) are present. By
changing this composition, either by shifting the mixing ratio
or the grain size, a wheel with tailored dressing behavior can
be realized.

d M ,70 + d M ,60 0,27 mm + 0,22 mm

d M 60/70 = = = 0,245 mm (4)
2 2

Furthermore, the number of grains per unit length (see

Equation (8)) is considered by calculating the distance that is
created between void space, bond and grain volume.
According to a perfect grain matrix, as illustrated in Figure 2,
results this

Figure 1: Real wheel indentation ind, according to the depth of cut and
forward feed [6].

The aim of the following calculations is to determine the real

contact length according to the angle arc at a0 dmax (see
Figure 2: Void space in a perfect grain matrix
Table 2). By means of Equation (1) the contacting sector
angle can be calculated and hence the real contact length in
Equation (2). Qi [4] reported that the real contact length in a distance of lgr=0,3348 mm. The gap between two grains
differs up to 175%. Due to the fact that previous equals lgap=0,09 mm and accounts for 26,8% of lgr. These
investigations have shown a much smaller deflection the distances are also further used for the calculation of L,M,max..
factor 1,1 was chosen in Equation (3). Equation (10)-(12) show the number of grains with respect to
cutting area, volume and time.
r a 63mm 0,5mm
= arcsin w 0 = arcsin = 82,77 (1) b 10,2 mm gr.
rw 63mm Nl = (1 l g ap ) = (1 0,268) = 31 (5)
d M 60/70 0,245 mm cm

90 7,2
l g = dW = 126 mm = 7,94 mm (2)
360 360 AA = lg,real b = 8,736mm 10,2mm = 89,1mm 2 (6)

l g ,real = 7,94 mm1,1 = 8,736 (3) 2

gr. gr.
N C , AA = N l2 AA = 31 2
0,891cm = 856 cut.ar. (7)
cm
720 A. Fritsche and F. Bleicher / Procedia CIRP 26 (2015) 718 723

gr.
3
gr. grains are part of L,M,max. In addition ind is subtracted, as
N v = N l3 = 31 = 29791 3 (8)
cm cm this indentation is the active cutting depth.

cm gr. gr. L, M ,max = d M 60,70 1,5 + l gap ind , (10)

N T = Vw N C , A = 2500 856 = 2,14 106 (9)
s cut.ar. s
L, M ,max = 0,245 mm 1,5 + 0,09mm 0,0014 mm = 0,4561mm
The total number of grains passing through the grinding area
is given by 2,14106 grains/s whereby only a fractional L, M ,max d M 60,70 0,4561 0,245
x= = = 0,1055 mm (11)
amount takes place in the cutting operation itself. This 2 2 2 2
behavior is treated via a statistical approach in the next
section. The mean value x respectively p describes the beginning of
active cutting grains and hence the active cutting area. This
leads to an area of 3,4% in Equation (16) according to Spiegel
3.2. Calculating the cutting area after dressing [7].

The presented statistical approach is illustrated in Figure 3 p = x

3,9
= 0,1055 mm
3,9
= 1,805 (12)
where the circumferential area is the area of particular interest L, M ,max / 2 0,4561mm / 2
and therefore splits up in three main layers, namely:

e
1. Grains with no contact layer (red): 1 x 2 /2
P( p) = dx 3,4% (13)
Describes the virtual border where grains have no 2
p
contact during the grinding process besides a
supporting and holding function
With respect to the sampling area of AE,A~4,48 mm and a
2. End of active grain layer (green): sample edge length of Sample.~2,156 mm in rotational
Describes the physical border which is the result of direction (see Table 3 and Figure 5) the number of active
the dressing operation, hence the active grinding
cutting grains can be calculated through Equation (18). For
layer
further comparison with the experimental results in section 4
3. Sacrificed grain layer (blue): Equation (20) provides the respective parameter.
Describes the physical border which is present before
the dressing operation and marks the abrasive wear gr. gr.
layer from the previous grinding operation N C = NT P ( p ) = 2,14106 0,034 = 7,27104 (14)
s s

These three layers on top of each other form the so-called N C , A = 856
gr.
0,034 = 29,1
gr.
(15)
Active Grain Stack (hereinafter abbreviated AGS). The cut. ar. cut. ar.
AGS from layer 1 to 3 equals L,M,max and builds the
enclosure for the normal distribution curve and the boundary
AG, A =
( 0,245mm )2 29,1 gr.
= 1,37 mm 2 of gr. ar. (16)
conditions concerned. Furthermore represents the dashed area 4 cut. ar.
under the normal distribution curve in Figure 3 the amount of
grains that fulfill the desired specifications. AE , A AG , A 4,48 mm 2 1,37 mm 2
AS = = = 0,068mm 2 (17)
AA 89,1mm 2

In contrast to the maximum possible number of grains (856

gr./cut.ar.) in the contact area of Equation (10) only a small
amount (29,1 gr./cut.ar.) is shown that fulfills the aspired
specifications. Expressed in percentages the mentioned
amount is reduced by 96,6%, respectively 3,4%. This number
of grains equals 0,068 mm2 cutting area at a grain cross-
section of 0,047 mm2.

4. Experimental investigations

4.1. Set up of the

Figure 3: Normal distribution of grains along the cross section of a wheel
according to the mean diameter dM 60,70.
Equation (13) therefore returns a distance of 0,4561 mm,
whereby the factor 1,5 takes into account that one and a half
The set up for the experiments is illustrated in Figure 4 with
the following components: (1) the grinding wheel with the
mentioned grain composition, (2) the work piece for grinding
experiments, (3) the diamond dressing tool with two diamond
 A. Fritsche and F. Bleicher / Procedia CIRP 26 (2015) 718 723 721

bars, (4) the copying sheet metal for capturing the wheel wear
after grinding and (5) the nozzle and pressure sensor for
cooling supply. The dressing tool itself is illustrated in the
right corner of Figure 4. One diamond has a rectangular shape
with the dimension 1x1 mm and a distance between the two
bars of 3 mm. The dressing operation is conducted via a
vertical Z-axis spindle movement with a feed of 900 mm/min
and a rotational speed of 3800 U/min (=25 m/s circumference
speed). Furthermore the infeed is 0,02 mm for each stroke and
equals 0,8 mm in total. By means of these parameters it can
be assured that the whole circumferential area is processed
and prepared for the graphical investigation (see section 4.2).
These error marks are filled manually with the next logical
adjacent color. The result of this revision is illustrated in
Figure 6, where the red color equals layer 2 and the orange
color layer 1 in the AGS. The blue color on the very bottom
indicates the total loss of grains and is not investigated
further. Besides these colors black indicates the walls of the
valleys, which appear due to radius deflection. In the middle
walls are hardly detected. They appear as the deflection
advances.

Figure 6: Processed picture for the area analysis with the three main heights.

Figure 4: Experiment set-up with grinding wheel (1), work piece (2), dressing The evaluation of Figure 6 results in Table 3 and hence a
tool (3), copying sheet metal (4) and cooling nozzle with pressure sensor (5). grinding surface area of 4,492 mm. The red area is of
particular interest because it indicates the probable cutting
area. According to Figure 2 AE,A is reduced by 26,8% due to
4.2. Graphical investigation void space and bond. This approach matches the result of this
particular sample, though closer related to the average out of
For the graphical investigation an optical 3D-micro- 50 images i.e. 72,9 %.
coordinate measurement system with a chosen 5-fold
magnification is used. This magnification gives a maximum
sample area of 6,125 mm. On average sample areas out of 50 Table 3: Results of the graphical investigation for a chosen sample
images represent an effective image size of 4,48 mm which
equals an edge length of 2.786 mm x 2,15 mm. By means of Conditioned Grinded
Area alternation
this measurement technology it is possible to gather height state state
information from top to bottom of each sample. Furthermore Grain
Area Spe- Area Spe- Area Spe-
the white error marks in Figure 5 show regions with either too AE , A cific AE ,G cific AR cific-
level
much reflection or an undercut with shading, thus containing area area area
no information. [-] [mm] [%] [mm] [%] [mm] [%]

Red 3,432 76,9 3,223 76,2 -0,209 -7,3

Yellow 0,668 14,8 0,993 23,5 0,325 7,2
Blue 0,013 0,29 0,015 0,4 0,002 0

Black 0,379 8,43 0,466 - 0,087 -

Sum 4,492 100 4,697 100 0,205 -0,1

4.3. Compensation if the radius deflection

As mentioned in the previous section the radius deflection has

Figure 5: 3D Alicona height profile after the dressing operation with error
marks (white spots) and black walls due to radius deflection. to be taken into account. Therefore Figure 7 illustrates the
applied approach. It is assumed that the second layer in the
AGS is evenly distributed over the sample area. Thus the error
722 A. Fritsche and F. Bleicher / Procedia CIRP 26 (2015) 718 723

is calculated by Equation (22) and gives a deflection of 9,92 Equation (24) shows the experimental sample area AS,E for
m. In order to determine the real indentation ind,def comparison with the statistical investigation result AS from
according to deflection, Equation (23) provides the respective Equation (20).
operation. The result of 6,4 m represents the depth of the
disc topography which takes place in the cutting process. AS , E =
4,48 mm
0,95 % = 0,043 mm (21)
100

200

-200 225,6 m 0,3%

0,95%
-400
208,1 m
-600
201,7 m
-800

Figure 7: Compensation of the deflection due to the radius of the wheel.

0 20 40 60 80 10

lsample / 2 1,0781mm
error = arcsin = arcsin = 0,980 (18) Figure 9: Load diagram of the sample area labelled with the real indentation
rw 63 mm
ind,def and the corresponding percentage.

( )
2
error = 63 mm rw2 lsample / 2 = 0,00992 mm (19) The comparison of these two figures divers 0,025 mm
(=0,068 mm - 0,043 mm), whereas 0,05 mm is the
measured cutting area out of 50 images. The remaining
9,992 m
ind ,def = ind , + error = 1,4 m + = 6,4 m (20) deviation of 0,018 mm can be traced back to the fact that the
2 2
grain distribution varies between different grain batches and
hence is either shifted to the right or left. In the presented case
higher percentage means a shift to the right side of the normal
4.4. Investigation of the height profile
distribution curve, thus a higher percentage of smaller grains.
The height profile of the chosen sample is illustrated in Figure
8. This graph shows the result of the primary profile across
the whole sample area with a maximum height of ~292 m
and matches the prediction in Figure 2 (lgr~average out of 50
images = 310 m). Referring to this knowledge it is assumed
that the grain layer is ~10% higher than the mean grain
diameter dM 60/70.

100

50

-50

-100

Figure 10: Normal distribution with 139 classes at a class size of 2 m.

-150

0 0,5 1 1,5 2 2,5 2,8

5. Alternative grinding wheel compensation

Figure 8: Mean height distribution across the sample area with the illustration
of a grain with the average diameter.
Expressed in percentage the real indentation ind,def leads to a
meshing area of ~0,95 % (see Figure 9). Starting from 225,56
m as maximum height minus 17,54 m (=3 Sigma=0,3%),
equals the final depth 201,754 m (=0,95 %). In addition, the
histogram in Figure 10 displays the distribution in classes
(class size 7 m). It can be seen that only a small amount
takes place in the cutting process itself (red bars). As long as
the wheel shows no wear only these 2-3 bars actually cut.
In Table 4 the respective compositions for the chosen wheels
are listed, whereas only the first three wheels (B181, B126
and F60/70) are existing abrasives and the remaining ones are
fictitious. The third and the fourth wheel only have the mixing
ratio in common to demonstrate the impact of the grain
volume onto the amount of active cutting area. Furthermore,
the fifth wheel shows the effect which would occur with a
single grain size. Finally, the mixing ratio is unevenly split in
order to cover all variations.
 A. Fritsche and F. Bleicher / Procedia CIRP 26 (2015) 718 723 723

Table 4: Percentage of bond, grain and pores in abrasive for the chosen This result confirms once again that changing the grain size
sample disks
does not affect the active cutting grain area as much as the
grain fraction. In order to cover all variations the mixing ratio
Grain Mixing Wheel in the last example F54/70 is unevenly split (30/70). It shows
Bond Grain Pores
Size ratio specifi- that the grain fraction is the same as in the F60/70 example,
[vol.%] [vol.%] [vol.%]
[mm] [wt.%] cation
while the other parameters are cut to 50%, as expected. In
CBN
B181 - grinding- 21 42 27 conclusion the grain distance lgr is affected to the greatest
wheel extend when changing the grain size or mixing ratio. In order
CBN to find the most suitable grain-dressing-cutting area
B126 - grinding- 21 42 27
specification, the pore volume has to be adjusted to the
wheel
AL- respective operation through practical experiments.
F60/70 50/50 Oxide 8 38 54
wheel 6. Conclusion
fictitious
F54/90 50/50 compo- 10 58 32
sition This analysis gives an insight on how different attributes and
fictitious parameters affect the dressing result starting with the
F90 - compo- 10 58 32 geometrical contact behavior and the Active Grain Stack
sition
fictitious layer model, to the abrasive composition and even to the
F54/70 30/70 compo- 8 38 54 statistical and graphical investigation. Especially the AGS
sition model is a good tool to describe the structure of a grinding
wheel and how the different layers interact with each other.
The following statements refer to Table 4 with the associated Furthermore the optical investigation is a suitable method to
investigation results in Table 5: evaluate the result of different dressing operations and to
proof the presented analysis. Practical experiments are still
Starting with the first wheel (B181) the FEPA chart for CBN inevitable but now it is possible to define a starting point
shows an average grain diameter of dM,181=0,174 mm. The where to hook up and optimize an already existing process in
corresponding grain spacing accounts for lgap=51,7 m. In a more efficient manner.
comparison with the wheel investigated in detail (see Table 1)
the active cutting area of this composition with 4,27 % is 18 7. References
% points higher. This increase as well as the fact that CBN is
much more wear resistant leads to a longer lasting wheel life. [1] Wegener K., Hoffmeister H-W., Karpuschewski B., Kunster F.,
Hamann W-C., Rabiey M. (2011): Conditioning and Monitoring of
Only by changing the grain diameter from 0,174 mm to 0,123 Grinding Wheels. In: Annals of the CIRP 60(2), 757-777.
mm, as applied in the second example (B126), the grain [2] Jiang J., Ge P., Bi W., Zhang L. Wang D., Zhang Y. (2013): 2D/3D
ground surface topography modeling considering dressing and wear
spacing lgap is reduced by 30 %points, whereas the active
effects in grinding process. In: International Journal of Machine Tools
cutting area increases slightly by 2 % points, i.e. from 4,27% and Manufacture 74, 29-40.
to 4,36%. It appears that by changing only the grain size a [3] Brinksmeier E., Cinar M. (1995): Characterization of dressing
significant increase in active cutting area cannot be realized. processes by determination of the collision number of the abrasive
This fact is implemented in the F54/90 example with a 50/50 grits. In: CIRP Annals Manufacturing Technology 44, 299304.
mixing ratio and a grain fraction of 87,6%. It can be seen that [4] Qi H.S., Rowe W.B., Mills B. (1997): Experimental investigation of
by increasing the grain fraction by 16,4 %points (referring to contact behavior in grinding. In: Tribology International 30(4), 283-
73,2% from subsection 2.3) the active cutting grain area 294.
doubles (6,55%).Applying this example to a composition with [5] Rowe W.B. (2009): Principals of Modern Grinding Technology.
the same parameters but at a single grain size, as in example Elsevier Science.
F90, the active cutting area almost stays the same (6,68%). [6] Malkin S., Guo C. (2008): Grinding Technology. Theory and
Application of Machining with abrasives. Industrial Press.
Table 5: Summary of the most important results of all investigated [7] Spiegel M. (1968): Mathematical handbook of formulas and tables.
wheel compositions. McGraw-Hill.
[8] Blunt L., Ebdon S. (1996): The application of three-dimensional
surface measurement techniques to characterizing grinding wheel
dM lgap gr.frac L,M,max P(p) lgr
topography. In: International Journal of Machine Tools and
[mm] [m] [-] [mm] [%] [mm] Manufacture 36(11), 1207-1226.
B181 0,174 51,7 77,1 0,3113 4,27 0,226 [9] Rowe W.B., Morgan M.M., Qi H.S., Zheng H.W. (1993): The effect of
B126 0,123 36,5 77,1 0,2196 4,36 0,159 deformation on the contact area in grinding. In: Annals. CIRP 42(1),
409-412.
F60/70 0,245 99,0 71,2 0,4651 3,21 0,344
[10] Brecker J.N., Shaw M.C. (1974): Measurement of the effective number
F54/90 0,237 33,4 87,6 0,3884 6,55 0,324
of cutting points in the surface of a grinding wheel. In: Proceedings of
F90 0,155 21,8 87,7 0,2529 6,68 0,176 the International
F54/70 0,125 50,6 71,2 0,2367 3,28 0,175