1. MRR77- 12
,:I t&Burring
1977
soclEiYOF
MANUFACTURING
ENGINEERS -
20501FORDROADd.
f!O.BOX93Q
‘DEARBORN -
‘MlCHItiAN,481i8
DEBURRINGBYCENTRlFUGALBARREL-I-UMBLING
LaRouxK. Gillespie
Senji,orEngineer i,’
Thb$endix Corporzkm q ” i
I
A&TRACT
Thereliability of smallprecisionmechanismsgreatlydepends
uponthe productionof burr-free, sharp-edgedparts.
Centrifugal barrel finishing (Harperizing)is oneofthe few
processescapableof producingtheseconditions. Burrs
lessthan 0.001inch thick by0.001inch high@&4x 25.,4pm)
can beremovedfrom 303 Se.st$inlesssteel, 1018steel,.and
6061~T6aluminum with dimensionalchangesin the order
of 0.0001inch (2.54urn)andfinal edgeradii of 0,003inch
(76,2-urn). Theseconditionscanbeproducedin batchlots
in 20minutes or less. Surfacefinishes canbereduced ~
’ from 45to 25or 35microinches (1.15to 0.68.or 0.89 Vim),
with 60minute,cycletimes. Stocklossesappearto be.
repeatablewithin _fO.O0006inch (1.524grm). Very small
parts receivelessactionthan parts0.5 inch (12.7mm)in
“diameter. .-
Cre&ve Manufacturing Engineering Programs
v
2. PubIlished August 1976
Prsjact Leader? 221
L. K. Gill. spie
Departmen' ,222
Project Team:
B, J. Neal.
R. L,. Kirkpatrick
C. E, Roebuck
MRR77- 12
.’
3. DEBURRINGBY CENTRXPUGALBARRELTUMELING_.
Prepared byeL. K, Gj,lLespis
c
The reliability of small precision mechanizers greu t Ey depends upol,
the production of burr-free, shasp-edged parts, CsntrlPugal
barrel finishing (Harperizlng) is one of the few processee capahla
o$' producing these condktions. Burrs less than O,.OU3.inch Thick
by O.OOL,inch high (25.4 x 25.4 /MII) can be removed fxxn 303 Se
stainless steel, 1018 steel, and 6063X-7% almlnum wi,tb diwensLonal.
changes in the order of (3.OOO;tinch (2.54 pm> and f:i.asi edge
radii uf 0.003 inch (76.2 em>. These conditions can be producW
in batch lots in 20 minutes us less. Surface finishes call bi;
?x?duced from 45 to.25 or 35 micrainches (1.35 tc~ Cl.68 or 0,89 @,J#
with 60 minute cycle times. Stock losses appear to be repea,tahLe
within 10.06006 inch (E.524 pm). Very small parts receive less
action than parts 0.5 inch (12.7 mm) in dlamstsr:,
This report was prepared as at? accounr of work sponsored
by the United StatesGovernment. Neither me Untted States
nor the United States Energ$ Aesirch and Oevelopment
Admlntstratlon, nor anv.of their emplo&es. nor any oi’their
contractors, subcontractors, or their employees, makes
any warrantv, express or Implied, or assumes any leg81 lia-
‘bility or responslblliw for the accuracy, completeness or
usefulness of anv information, apparatus, product or process
disclosed. or represents that ita use would not infringe pfi-
vatelv owned rights,
_’
.MRR77.-'12c9
4. 4
t
-5
Sectioo. *
SUMMARY 2 l * . e / P - .
DISCUSSION, f 1 / 1 . s L
SCOPEANDPUlZSlOSEe . .
PRIOR WQRK.. . . . . .
ACTIVITY. . . . . . 1 .
CONTENTS
Pa@? F
. . *
. . I
. . .
. . -
. 1 3
. . .
. . .
. . II
* . *
. 1 .
‘ . /
1 I .
. I .
. . .
. . *
. . I
EfQec$ of the ProceSs on Edge Radius, Sz&-
aad Surface Finish. . . . . . . . .- . + / .
Repeatability of Stock Loss . . , . . e . . .
Effect of Burr Size ga.I$dge Radiui. . ‘g . . .
Observations on Production Parts. . . . . . .
Effects.of Piece Part Geometry~. . , . . . . .
Barpesi.zing Effects on Subsequent Processes .
Evaluation of Published Literature.' . . . . .
ACCOMFLISHJdENTS... . . . . . . . . /) . . . . .
FUTUREWORK. . .'. . ; . . . . . . . . . . . .
REFERENCES., . . . . . , , ': : . . . . . . . . .
APPENDIX A.. GUIDELINES FOR USE OF CENTRIFUGAL
BARRELTUMBLING. . l . . . . . . . .
DISTRIBUTION, . . . . . . . . . '6 . . . . . . . .
i
. i
‘ L
I .
. .
” I
. .
. .
. .
I I
1 -
. .
. .
. .
. *
. .
. 4
, . ‘.
. . .
8
e
9
I.0
I 3
43
47
53
54
57
62
., 62
62
63
64
82
,.:
,
MRR77-12-
-2..
5. -4-
ILLUSTRATIONS
. Figure-,
-I " Schematic of Centrifugal Ba.rrel Finishing
*
L
2
3
4
Unit * a.. s * . l * . . . . . . . .-* . ”
Typical Par%icles Used on Preci:;;~on IMinLaturo
Parts (P=939%1). . . . . . + t c . I . . .
Cross-Sectional View of Typical Uurr _ . . "
Haspesizing Effects on 1018 Wee1 Cylinders,
Using Nl4 Nuggets and IA-1 Compound at
15 g's . ,. . . . . . . * * I . . . . " ‘ .
5
6
7
/
8 , i
,’ Ii ’
9
10
3,'11
s
* 12-
13
:', ,<.
Harperiaing Effects on Stainless Steel
Cylinders, Using N14 Nuggets and lA-1
Compaupd at 15 g's . + . + . . . . . , e .
'arperizing Effects on Al.uminum,:~yLinders,
r Using N14 Nuggets and IA-1 Csr&ound at
: 15 g's . . * * . . * . . . . . . . . s . .
Harperizing Effects on 1018 Steel. Cylinders,
Using Nl.2 Dolamite and No Compound at
J5g's.. l . . . . * . ..I). l . ..(.
Harperizing Effects an Stainless Steel
Cylinders, Using Nl.2 Dolamite and No
Compound at. 15 g's . . . . . . . . . . . .
Harperiz+ng Effects on Aluminum Cylinders,
Using N12 Bolamite and No Compound at
15g’s. l I . . :. . . . * . . . . . ‘. .
Harperizing Effects on 1018 Steel Cylinders,
Using 3/16 Triangles and lA-1 Compound
at 15.g's; ; . . . . . . . . . . . . . . '.
Harperizing'Effects on Stainless Steel w
Cyfindetis, Using 3116 Triangles and lA-I
Compound at 15 g's , . . . . . . . . . , .
Harperizing Effects on Aluminum Cylinders,
Using 3/16 T'riangJes and lA-1 Compound at
15 g'sY , : '. , ;; . l . . . . . . . . . .
Harperizing Effects on 1018 Steel Cylinders,
Using N24 Nuggets and lA-1 Compound at ~
15 &ti- . . -I: . . . * . . . , , , . l * , ,
MRR77-12
.- :"
i
Page
I
3.Cl
I.7
28
14
20
. 21
. 22
. 23
. 24
. 25:-
t ’ 26
6. ..”
-5 c
14 Harperizing Effects on Stainless Steel
1 Cylinders, Using N24 Nuggets and IA-1
Compcund at 15 g's v j 5 s i + s . . * + s I
15 ‘/
16
17
18
19
20
21
22
Harperizing Effects on Aluminum Cylinders,
Using N24 Nuggets and LA-l Cclalpolind at;
15g's.;. . a., *, *. . . / *. . . ‘
Harperizing Effects 0n 1018 S’taol. CyI.i.ndc?rs,
Using Nl4 Nuggets and IA-1 Compound at;
2Og's l , i. i i . . . . . ...* . . “.
Harperiaing Effects on Stainless Steel.
Cylinders, Using Nl4 Nuggets and IA-L
Compound at 20 g's . . . . . . . . w . . . .
27' "
Harperizdng FPfects on, Aluminum Cylinders,
Using NliJ: Nuggets and IA-1 Compound at
20 g's . . . . . . . . . . . , . . . . . . * ,_..:1
Harperizing Effects on 1018 Steel Cylinders,
Using Nl4 Nuggets and lA-;i Cor@ound at
5 g’s. . * . . . . . * . . . . * l . l i 1 + 32
Harperizing Effects on Stainless Steel
Cylinders, Using X14 Nuggets and IA-1
Compound at 5 g's: . . . . . . . . . . . . . 33
Harperizing Effects on Aluminum Cylinders,
Using N14 Nuggets and lb-1 Compound at
5gts.../, . . . . . . . . . . . . . . f 34
Harperizing Effects on 1018 Steel Cylinders,
Using k[4-Inch Plastic Pyramids and l,A-1
Compound at 15 g's . . . .' , . . . . . . . . 35
23
24
Harperizing Effects on Stainless Steel
Cylinders, Using l/4-Inch Plastic Pyramids
and lA-1 Compound at 15 g's. . . . . . . . . 36
Harperizing Effe$ts on Aluminum Cylinders,
Using l/4-Inch Pla&ic Pyramids and lA-1
Compound at 15 g's . . . . . . . . . , . 3 . 37 L
25 Harperizing Effects on 0.5-Inch (12.7 mm)
Stainless Steel Cylinders, Using Various
Media aad lA-l Compound at 45 g's. . . . . . 38
a
26 ,Harperizing Effects.on O.J-Inch (12.7 mm)
.,lY Stainless Steel.Cylinders, Using N14 Nuggets
and lA-1 Compound at Various-g Levels. . .,. 39
a
1 .’
MRR77-12 .:
7. .,,,..‘I’,‘.
-6-
27 Conceptual View of Change in Burr Height . . 4
. 28
29
.
30
Redugtion in Burr &i&L by Harpwizing. , , ,
Miniature Stainless Steel Dl0c.k~ Df3iorfj km1
After Barperizing (P-91669)* . . . * * . " e
31 Stainless Steel Pinions Before und Afte:x+ I
Barpesiziug (P-91670). . . . . , . , . . , ,
32 Aluminum Spacers &fore and After Lfarperizing
(P-90771). , . . . . . ‘ " , . I - I ‘ . 1 .
33
34
35
36
Powder Metal part. . . . , , , . , . ,* . , . ,
Effect of Part Geometry on Deburring , , ti j/ ,
Fffect of Geometry on Edgo Radiu,gging , , . . ,z
Effect of Edge Angle and Vibration Ttie on
Edge Radfusiag of Phosphor Bronze
Workpiecse. . . . . . . . . . . . . . , . . .
A-l
A-2
A-3
Mediakodged in Slots (q-93289), . . . , , . .
Media Size Variation in Dolamite (P-90774) . .
Illustsation of Edge and Corner Break
Definitions, . . . . . . . . , , . . . . . .
_I
.
-
MRR77-12 .’
8. -7-
TABLES
40
4.1
44
45
46
'i 51
64
66
67
77
78
79
.
80' -
i
Number .
' Specimen Size , . . . , . , . , . . .,. + , ,1
2
3 Equations for Edge Radiusing, 30:2 Se
Stainless Stuel . . . , , , , , , . . , , .
Equations fo; Edge Rztdiusing, 1018 Steel.. . .
Equations -Ec?rEdge Radiusing, 606X-7%
Aluminum, . . . . . . . v , , , . . . , . . e
4
5
Rates of Change in Radius (Microincb~s per
Minute)...............,..
Time Required to Produce a GiveneRadius , I ,
Repeatability of Siize Change. , , . . , . , .
6
7
8
9
A-l
A-2
A-3
IMeanChange in Burr Height. . . , , , . . . ,
Description of Media anti Abrasives Used . . .
Typical Weights of Abrasive Media . . . . . .
Selecting Centrifugal Barrel Finishing
Conditions. , . . . . . . . . . . . . . , ,
Typical Media Size Variation and 'Role Sizes
in Which Media Will Lodge . . . . k . . . .
A-4
A-5 Criteria to Use to Screen Media . . . . . . .
A-6' Burr Sizes Typically Found on Precision
Miniature,Parts at Bendix . . . . . . , , .
A-7 7 Conditions,Which Should Result in Complete
Burr Removal While Maintaining an Initial
Surface Finish of 32 Microinches
(0.813 Pm), Stainless; Steel Cube,
0.125 inch (3.175 mm) on a Side . . . . . .
MRR7.7- 12
9. I
-8-
* SUMMARY
.
Component parts of small mechanisms need near-sharp edges to bf2
- reliable; parts must also be burr free to avoid jumming the
mechanism if burrs should break loose during operation. J.rj, 'the
past machining burrs have heen removed by hand bscauae al: these
needs, but hand deburring is time-consuming and inherently
operator-variable.
This study was initirtted to determine how centrifugal barrel
tumbling affected the surface finish, edge radius, and dimensicins
of precision miniature piece parts. The study involved 5800
measurements on bath test specimens and actual prndup,tion parka.
Far normal conditions, burrs O.QO1 inch thick by 0.001 inch L1
high (25.4 by 25.4 pm) were,-removed in 2 to 5 minutes. Thickeri '
and higher burrs required much longer times. Et is possible to
remove burrs 0.001 ihch or less while removing only 0.0001 inch
(2.5% pm) from external surfaces. For zz$Ller burrs, complete
burr removal can be obtained with stock'l%sses of 0.000050 inch
(Ii.27 pm) or less. The repeatability (standard,deviatian) of stock
loss on many samples was within 0.000020 (0.51 pm).
Nominal edge radii of 0.003 to 0.005 can be easily produced with
this process, but smaller radii can be produced only when burrs
are O.OOE inch thick or thinner. .Typical burrs found on miniature
parts produced from 303 Se stainless steel, 1018.stee1, and
6061-T6 aluminum" are 0.003 inch thick by 0.003 inch high (76.2 by
76.2 j.rm). 4'.
For the conditions studied,in this test, surface finishes improved
from 45 microi#ches'(X.M pm) to 25 microinches (0.68 pm). Surfice
finish and radius changes are exponential functions of time, ~?ri?i!l.e
stock losses are typically linear. High velocity (high g levels)
increa'ses abrasive action greatly, as does the use of large media.
The effectiveness of deburring is an exponential function of
burr'thickness and running time. Thick burrs are removed much
slower than thin burrs. WhiSe the process is fast, works well
* on miniature parts,, and is very repeatable, the conditions used
must be carefully matched to ,part geometry, size, and material.
Very small parts receive much less action than larger parts. r
c
MRR77-12 .'..*
10. DISCUSSION
t
SCOPEi.iW PURPOSE
PRIOR .WORK
No prior' studies of centrifugal brtrrol tumbl.ing have beon ri2pGrtc,d
by Bendix, although three related
vibratory deburring."'~"
studics have been r:c?porl,r;cl r’jn
ACTIVITY
Centrifugal barrel tumbling, as the ncameimplies, j-s the process
of allowing parts to tumble in a rotating barrel under a centrifuga3.
force. A centrifugal barre$ uni$%s muchzllke a ferris yheel,
except the barrels rotate in the opposite direction of the outer
portion of the machine (Figure 1). The rotation of the turret to
whidh the drums are attached creates the centrifugal force. The
barrel rotationLproduces a continuous sliding action of parts and
abrasive within each barrel.
Like all loose abrasive deburring operations, parts are mixed
with loose abrasive particles, an abrasive compound, an,d water.
The particles, which aqe,much like sand pebbles, flow over part
surfaces removing burrs and polishing the surfaces. Some typisal
particles used on precision miniature parts areshown in Figure 2.
Precision as used here indicates that piece part tolerances are
less than 0.002 inch (50.8 pm). Miniature is used to indicate
that the largest piece part dimension is less than 1 inch (25.4 mm).
The majority of parts and applications described in this report
are for parts roughly 0.25 inch (6.35 'mm) in size or smaller
'having at least one tolerance less than 0.001 inch (25.4 pm).
These abrasive particles are available in over 500 combinations
of sizes, shapes, materials, and degrees of aggressiveness.
Thi-s process has several advantages over other processes for
.
debitrring precisioli mi:niature parts.
cc It is a mass finighing operation (that,is, it deburrs many - .
parts in a single cycle).
0 It has very short cycles (10 to 20 times faster than vibratory
deburring).
0 It is specifically designed for small parts.
b
MRR77- 12
11. Ivgure 1. Schematic of Centrifugal Barrel Finishing
Unit j
0 It is a very flexible process (easily adaptable to different
part shapes, s$zes, and tolerance*'restrictions).
r) It cleans and polishes while removing burrs.
General Test Details
Three basic tests were performed in this study. In the first,'
cha,nges in diameter, surface finish, and edge radius were monitored
for seven different conditions, three part sizes, and three work-
piece'materials, In the second test, the repeatability of size
change was evaluated for 'two materials. In the third test, a-n
attempt was madeto quantitatively evaluate.how burr size influences,
the,edge radius produced.
In all tests a model 2VM-13 centrifugal barrel tumbling machine
(Harper Buffing tiachine Company, E&t Hartford, Connecticut) was
used. Although Soviet;, Japanese and other manufacturers have
recently been introduced commercially, centrif gal barrel tumbling
is often called Harperizing because the p$oce 8s-was developed by
and' the majority of units available in the ,United States were,'
I
MRR77- 12 .
.’
-
*
- .I
-
12. ..-z- 1 / . “~._ a. _-.
; I’
.’ .
-ll-
I,r
,
&
*
Figure 2. Typical Particles Used on Precision tiiniature, Parts
produced by the Harper Buffing 'Mai=hline Company. The-term Harperizer 'I
will be used in this report synohomously with centrifugal barrel
tumbling.' This mach$ne is horizontally mounted and utiliz & s
13-inch-diameter (330 mm) barrels. 'l&rThese barrels hold 2 qu.ts
(2.2 liters) of abrasive particles (media) and typically up t&,
1 quart (1.1 l+iter) of parts. As seen in Figure 1, the machine.' *
has two barrels located at distance ++R++from the center of the
turret. The distance R for this pachine is 12 inches (305 mm).
Centrifugal force is expressed by:
'mV2
'FCJ3NT= -if- =
4WR?T2112
38OOg ,(l) ;
1
(1 MRR77-12
13. where
m = Mass (mass of all objects in barrel in tkS.s ease);
V = Veldcity of center bm barrel;"
F= Centrifugal force;
R = Radius of rotation;
w = Weight (of contents of barrel in this case);
n = Numb&- of rehlutions; and
GE= Acceleration of gravity.
In the English system of units wfrere W is in pounds, R in feet,
and n in RPM, Equation 1 reduces to
FCENT= 0.000341 WRn' ..
For a radius of 12 inches (305 mm), a barrel content weight of
I.0 pounds (ten percent of which is assumed for illustration
purposes to act on the parts), a speed of 209 RI%:) the parts in
the machine are subjected to a 15 pound grinding and deburring
fOX-33 C
For ease of calculation and machine control the force level is
generaLly described in terms of ++g++forces (i.e:, the force is
I 'IX++times larggr than the force of gravity). Thus, L
mV2 1FG= R
mVE
(3)
I 0.000341; Rn2g (in English system of Units). (4)
I ' .,
where FG is a number representing how-much larger the force is
than the force of gravity. For the,values given abbve, the
machine produces a force of 15 g+s. Throughout the remainder of
.this report' forces will be described in terms of g forces.-
The above analysis is only approximate because it is ,basad on
the center of the barrel and the -parts’and media.slide along the
inside diameter of the barrel, a. dist*ance, rb f.rOm the Center.
Metsunaga 4+5rS* hasdeveloped more accurate equations to describe
actual for&w and motion of the parts. Because each model of
centrifugal’barrel units, has. ‘diff,ereat ‘vai’ues ,of e and rb, the :
results of the studies in this report will not exactly reflect
th? results ,to be found in other machines.
I,a 0 : I
;c *’
,*
.,.' MRR77-'12
14. +tiiCCtlly, the machipe was operated by placing 2 quarts (2.2 liters)
of abrasave media in each barrel, adding a cup of abrasive compound,
inserting parts, and covering the entire mixture with water.
_After running for a specified time {ZO minutes unless otherwise
noted) the -barrel was flushed with clean water, a burnishing
powder‘was added, ,the machine was cycled for an addi.tl.c,nu.l
5 minutes, and then rinsed. Pasts were then @:eparnted from the
media. More complete details of the procedure u:jed tire given j.r~
Bendix Process Engiineering Speclficatzian (PES) P-1251048. Tht;:
media and absasiv i.4compounds used are described in &-t&U itl'
Appendix A of thia report,
Effects 09 the Proce~+.on,.,cj.i.e El- Size, and Surface Fin.ish-
In this test, solid cylinders of 1018 steel (RB93), 303 Se SI:a9,ni,~:,;3s
Steel (RC21), and 6061-T6 aluminum (R@Q) were subjected to
various combinations of media and g forces. The edge radius,
change in diameter, and surface finisb were recorded five times
fn a total deburring cycle of 60 minutes. Each time the specimens
were reinserted in the barrels, the barrel was flushed and fresh
iabrasive compound was added, Table 1 indicates the three si.zes
of specimen used. The diameter of each sj%cimen was measured to
the nearest 0.000020 inch (0.508 pm>. To assure close uniformity,
all specimens were originally centerless ground to 40.0001 inch
.(;r2.54 pm> and cut to length in a monoset tool grinder using a
0,015 inch (381 pm) thick abrasive cut-off wheel. This wheel
typically produced burrs of the size shown in Table 2, iYlost
screw machines as used at Bendix produce burrs of this same size.
Most machining processes produce burrs two to three times larger
than.these values. The definition of burr properties is shown in
Figure 3.
Ten specimens were used of each specimen material and each size
in each of seven tests. Edge radii and burr size were recorded
to the nearest 0.0001 inch (2.54 pm> using a Leitz optical
measuring machine. Surface finish was recorded to the nearest
microinch. The results of these measurements were averaged and
this average was used to plot the curves shown in Figures 4
thrdugh 24, Note that two readirigs of edge radius were taken on
each part which provided 20 readings at each combination. Mqdia
sizeqused are shown in Figure 2 .and described in detail in
Appendix A. Initial .burc height is plotted as 'a negative radius _
in Figures 4. through 24.
As Cseen in Figures 4 through 24, edge radius increases with time +
in an exponential f ashfon. In the majority of cases edge radii
do not exceed 0.010 inch (254.pm) after 60 minutes. A O.-O05 inch
(12'7 pm> radius occurs a$?ter roughly 20 minutes. Edge radii of
only 0,002 inch (50.8 pm) can be produced &fter a 10 minute cycle
under some conditions,
,The large part-s had larger radii than other sieti*imens run tinder
identical conditions.. In Figure 4, for example, after 20 minutes
MRR77-12 - .
I.
15. ‘.; ;’
Table'l. Specimen Size
-,
-T4-
0.490 12.446 0.500 _ 12.700
a.240 6.096 0,250 63.350
Q.115 2.921 6.125 3-175
in N14 nuggets at 15 gls, the l/8-inch (3,175 mm) specimen had a
0.003 inch (76.2 pm> radius while the l/Z-inch specimen had EL
0,QOG inch (152.4 om) radius.
Diameter changes foalowed a similar pattern of size dependency.
For the conditions just described, after 60 minutes the diameter
of the largest specimen decreased 0.0008 inch (20.3 pm) while the
small specimen Iost only 0.0004 inch:,(10.$5 pm). These stock
losses were roughly linear with time. 6
As a general rule surface finishes improved from 40 microinches
c(1.016 pm) to 24 microinches (0.610 pm) in 60 minutes. Surface
%i~tnEsh results, however, varied ,significantly with workpiece
mater&al and conditions used.
The aggr&s;tiyeness of the Harperizing action was roughly propar-
tional to the size of media used (8igure 25) and to the magnitude
of the forces used (Figure 26). !!
Ey extrapolating the data, it appears that the burrs were entirely
removed after roughly ? to 5' minutes in th?,, Harperizer. As
indicated later, thicker and higher burrs require more time far .-
complete removal. For such small burrs it,is possible to assure
complete burr removal with Xess than 0.0001 inch (2.54 pm) change
in stock,diameter. In some cases it is.possible to assure
0.900050 inch (1.27 pm) or less size change.
The data'in,Fi&res 4 through 26 tended to be highly repeatable.
. Edge radii, for example, had a standard deviation of 0.0003 to
0.0006 inch (7.6 to 14.2 pm). Diameters had a standard deviation
of 0.00002 to 0.0002 inch (0.5 to 5.1 pm). However, surface
. finjsh standard deviations varied from 3.0 to 13.0 microinch
,(0.0762 to 0.330 pm)., :
The data for radii.in Figures 4 through 24, was fitted to a
mathematical model I
,
R = ao + alt ,+&a2t2l (5) :
MRR77 - 12 . ?I-
, 1
16. . ,-
- . .
-15-
Table 2. Properties of Burrs Produced With
Abrasive Cut-Off Wheel
1
Workpiece
Material
Aluminum
0,5 Inch (12.5 mm) 1.2 (30.48) 2.3 (58‘42)
'0.25 @ch (6.35 mm9 1.3 (33.02) 2*3 (58.42)
0,X25/ Inch (3.175 mm) 1.3 (33.02) 1.6 ( 4 0 . fix )
0.25 Inch 22.86) 6.7 (17.78)
0.125 Inch 2: I 5*08) 0.3 (7.62)
303 Se Stainless Steel
0.25 Inch 0.8 (20.32) (10.16)
0,125 Inch 0.6 (15.24) 0340.3 (12.40)
*Values shown are average readings,
where
R= edge radius and ..
t = time in Rarperizer,
In fitting the data,, an 'attempt was made 'to define the initial
burr height as a‘negative radius. The fit was poor. The same is
true when an initial radiu.s of zero was assumed. Th'e quadratic
model has concavity either up or down 'when the plot of the data
shows concavity to the right. The model
R = at,b,I s
b3> ‘.
using the negative'radius when t = O'was tried,, also using R = 0 r
when t = 0. This gave very good resu&ts (the correlation coeffi-
"cients were typically 0.97 or better). The burrs on these-parts
,were relatively small and were removed rapidly. The predicted
radius at the end of l.minute exceeded the initial burr height.
Had the burrs been 'large, their size would have had to be considered.
17. Figure 3. Cross-Sectional View 0P
Burr
(HE I GHT)
Typi.ca..L ,
Tables 3 through 5 gives the fitted equst,ions far each process r~z‘
group, material, and sfze. The standard%rror of the estimate 'I~
is given for each. The burr at t = 0 was omitted for these
calculations (that is, at t = 0, R = 0). This fictitious point
was not u,sed en computing OR.
The fitted equations for predicting the radius at time "ttr minutes,
are valid for time intervals,of 0 to 60 minutes.
The data for the 20 and 40 mifiute intervals for Groups V and VII
indicate possible erro.rs in timing, Tha 20-minute time appeased
.to be more nearly 30 minutes and the 4O-minute time probably was
not run at all. The direct comparison of size and material 'I
effect within the'group is still valid because they were all
processed together. The standard error UR for Groups V and Vfl
reflect these consistent shifts for.20 and 40 minute time intervals.
An e$timate of. the radius after 1.0 minute harperizing (t = 1) is
I' the coefficient "a" in the above equationd. Group, III has the
largest coefficients for al.1 equatio.ns, indicating it is about
twice as fast as Group I, two'to three times as fast as Group II,
q and about four time-s faster than Group IV in forming the radius. _
A better estimate for the rate of change of radius per minute in
I the interval 0 to 60 minutes may be.obtained from the derivative
of the fitted equation'
dR La b tb-l
dT-
,
Text continued on page 43. .MR,R77-12
18. -17-
* 40
l 40 -
l 40
l
l
(VV H3NIOti3lW) HStNt,.A 3%Jtif’tS
MRR77-12
20. 0
1 I -+1y I
e 40
l l
0 Q Q
--
#=
l
l
0
MRR77-12s. . _
-
21. ._ f
-m-
c
l 4 o-
(HJNI) a33naotld snlawtl ",,,
c-4 -a-‘a2
-Y.--x
---
-40
c t""') HSlNlzi 33WJI0-tS
0
l
5: s
0 0 0M z
(WV H3NtOtl3lW) HSlNlf 33WMlS
MRRf7- 12
22. (w") 033Ilaotld snlawti
”
, I,.,.. ,.. :e:i,“““’ .L:,.i.. :
-21-
040
04 0 *
040
0 ln 0s: ;j z z
00 ri
d d d d
4
*. (~3Nt) amaotid snlawtl
a
l
1
0
I t I I 1
=: 2:. z
0 a 0
(WV H3N1013lW) HS1iId 33WMls
MRR77-12
23. Q 40
,’ I,
(HJNI) 3SW3UJXl bELL3wt110
(““1 a33flamd sn i am
l 40
C-lYca
. . .
e-1
(Qrl) HSiNIJ 33W~kKiS
l
MRR77-12 ---
24. ..I‘-
,r:_I ,- . I
-.
/
l 1
w
(~3~1) 3sw3ma t1313wwIa "
-23-
.,,
,
* l 0
I I I. ‘I
I,:
,.,’
:!
/ ,:
a
‘/,
.
2;
‘. ” l
25. _
(w? amaoud Sntawtl
r--
.
**a - l
bl * 0
I f I I
2
0
* ifi
0 0
(WV H3Ni0)131W) HSINI;I 33W$MS
(H~NI) cgmaoad smaw8
MRR7742
26. e 4 0
* 4 0 s
l 40
) a33naotld sntavu
(‘“u”) HSlNlj 33VJllfiS
l
l
hARR77- 12
27. .(. _
-26-
. a
C-l
0
8
I I I I
l 40
l ,4b
-0
.
l
w
.
d 0 0 a a
Ln a- m s -
(WV H~NIOWIW) HSINIJ 33wtfns
(~3t4 1) a33yotld sn i avtf
MR.R77- 12
43. II II
fz offi u$ ,a
. .
“V
nl-l r=. +J
-x -cdI
4
MRR77- 12
44. -43: -
The data in Table 6 are the rsttes of change of r~.~dettsper minute
at 5, 20, and 30 minutes.
Although"these rates are valid for comparison, the estimated I;irrle
to obtain a 2.0 or 5.0 mill radius from the fitted aqtlratians will
show processes are too fast and would be diQficult; lx control,
while others are too s:Lcaw, requiring more time than would be
uesirable. Values greater than 60 ini.~~ut(;tis am ex1;Capr’,.iai;isri rtnci
their accuracy will decref.~se with :ini:rettmd proj oci; iah I-,eyond
60 minut~,s 5 The plot'tecl curves i F-i.gure 4 through 24s muy bc$uuc:d
to esttiate the time to obtain a desired rad.ius if f.1; j.s obl,;~lnr?cl
ia 60 minutes or less.
Tables 6 and 7 show that the procevs for a given rnd.i.u~ part-,,
material, and' part size must be chosen cara*fulLy. h proc@s:; I;k~al;
will yield a 0.002 inch (59.8 pm) radius in a desirable time may
require an excessive time to generate a 0.005 inch (127.0 flm)
radius. Conversely, a process that gives a reasonable time to
generate a 0.005 inch radius may be %oo rapid for good control o41
a 0.002 inch radius for the same size pa& and material. The
change in time for a change in radius ,is not the same ratio for
different size parts, materials, or processes. The time increases
with a reduction in size or hardness of the material. The size
effect is not the same ratio for all materials. Group VII is a
fast process, second only to Group III. Group Iis the third
fastest. Group VI required the longest time to.obtain 0.002 and
0.005 inch. An attempt was made to express the values for rra11
and "b" in the fitted equation in termsof material hardness and
specimen size for each'process, using a linear model, The fit
was poor. A more flexible model, requiring more than three
levels of hardness and three specimen sizes, is necessary and
hence is beyond the capability of the existing data.
Repeatability of Stock Loss
In the‘second te$t, 30 each'of the O.S-inch-diameter (12.7 mm)
cylinders of aluminum and stainless steel were measured,(beforo
and after Harp-erizipg. As before, results were recorded to the
nearest 0.00020 inch (0.5. pm), In this instance, 'the part was-
rotated to find the minimum dimension at the center of the part.
(In the pre'vious tbst no attempt was made to find the minimum
reading.) The parts were measured twice after Barperizing to
establish the measureme,qt error.
The specimens were Harperized at 15 g's in N14 aluminum oxide
nuggets and U-1 compound for 20 minutes,
.KP +
.
.
c
4
MRR77- 12
45. I 5
20
30
II 5
20
30
III '5
20
30
IV 5
20
30
v 5
20
30
VI -
2;
30
VII 5
i::
218 I.57
103 70
,a3 56
152
74
60
377
158
122
147 89
72 46
58 38
181 128 80 228 156 92 210 162 109
82 82 73 101 '79 40 109 77 50
65 p/2 71 79 65, 32 90 62 39
255
7948
295
I.20
92
88 57 157 108 72 152 114 51
45 34 71 49 38 76 52 l.6
37 29 49 39 31‘ 62 41 12
212
85
65
579 479 381. 207 588 442 3C)Gi
75 203 166 84 235 3.92 124
58 158 130 05 qp 15J,. 86
37 Il.90 PO2 60 189 96 72
22 89 52 33 78 38 30
19 71 43 28 60 29 ~23
89 320 250 ~112 435 287 180
39 127 103 43 178 112 69
30 9i 80 33 137 85 51,
,
*1 Microinch = 25.4 nm.
The results of this study are shown in Table 8. The average
stock loss'for aluminum was 0.000072 inch (1.83 pm) and for
stainless steel it was 0;000049 inch (1.24 pm). From the statistics
shown, one can be 95 percent confident that 95 percent of all
losses produced on stainless steel specimens under identical
conditions kill fall within the Limits -0.000049,'+-0.000070 inch
(-1.24 kJ..78 pm): *w
46. Table 7. Time Required to Produce a Given Radius
II 0.5 8.40
0.25 111.75
0.115 196.56
rrr OS5 0.40
0.25 0.59
0.115 3.09
+ IV 0.5 8.65
cl,25 25.42
0.115 100.70
v 045 4.40
0.25 15.20
0.115 26.00
VI 0,s 15.20
0.25 26.20
0.115 54.20
VII 015 0.52
0.25 .1.40
0.115 21.20
3.75 0.24 2.70 0.07
8,05 0.57 5,60 0*40
35.58 1.65 22.56 0.34
58.35 4.37> Q33~48 2.31
143.21 20.11 117.14 f2.86
452.26 52,14 260*23 36.13
37.50 2.30 21.00 5*10
58.80 8.40 51.10 8.40
69.50 21.40 188.40 14.30
66.20 8.00 52.20 9.40
156.00 15.00 121.00 13,20
238.00 38.20 210.80 26.90
9.70 0.36 5.60 0.21
20.30 1.10 13.90 0.40
213.10 7.40 133.00 1.10
*O.l inch = 2.54 mm.
.%
The amount of stock lost was a linear function of' the size of the . "
specimen. The equations defining the loss are
_
DA
- 0.48900 = 4.4294 x 1O-5 + 0.7724(Di - 0.48900) , ( 8')*.
for aluminwn'
a
MRR77- ,12. ,+“..A
47. :_:--_ r >
-#-
Table 8. RepeatabFXity of &ze Change
Without Measurement -72 1268 -49 rn71)
Error (-1.63 kG.8j (-1 .24 21.72.l)
13A- 0.489’00 =c -9.2744 x lcr5 -I- l.0652(Di - 0.48900) w
for stainless steel
.I $j!i
II'
where Di is the initietl diameter and DA is the diameter after
IEarpGwizing. i.
These equations reduce to
DA = 0.111242 -I=0.7724.Di
for aluminum and
DA * -0.03179,5 + 1.0652 Di
‘(10)
w>*
for stainless steel'.
Note that these equations are only valid for diameters in the
vicinity .of 0.489 inch.(12.42 mm), lengths of 0.5 inch (12.7 mm)
and the Harperizing conditions described. Suitable conversions
A must he made if diameters are expressed in metric dimensions,
In many cases, the standard deviation of diameter measurement of
1 a group of parts after Harperizing was smaller than initial group
standard deviation, resulting in more consistent parts. This can
be explained'by the linear trend described above. The process
takes more.stock off a large part t,han one slightly smaller.
a
--
MRR77- 12 .I
48. This&can also be shown statistically. If a group of parts haviag
a stikndard deviation of OA are subjected to a process which also
affects dimensions in a random manner, then the standard deviation
of thegroup of parts coming out of this pracess will ba L
“2 2
UF = UA =+ tYB - 2raAuB
where
5 = final standard deviation of the group,
%I = final standard deviation of the process,
OA = final standard deviation of initial parts, arid
r =: correlation coefficient 0% before and after measurements,
In a truly random process, the before and after measurements are
not correlated; thus, r = 0 and UF will always be larger than GA.
The linear trend des~cribed above is a direct inQcation that
correlation exists. Therefore,
smaller than GA. As a result,
if r.,is pasitive, UF can be
the final %arts can be more
consistent than the initial parts. Note that the basic size will
still change, but there will be leas variability in the measurements,
As an example, from the data of this test, the initial standard
deviation in aluminum was 253.727 microinches (0,643 pm); after
the process, o~'was 235.807 microinches (0.599 pm) and r was
0.8311. Substituting these values into t&e previous equation, it
is found that the standard deviation of the process under these
conditions is either' 400 or 22 microinches (1.016 or 0.055 pm);
the solution to the quadratic equation ha$ two positive roots in
this case. Based,on a npber of other stu$es bf actual piece
parts it -is app,arent that ,the,repeatability (a~) is on the order
04 22 microinches.
The observed effect of -decreasing size. variability has been
uetected on about 50 percent ofr.all batches studied. Roughly
25 percent exhibited a slight increase in variability and the.
remainder exhibited no change. In most cases, however, the
improvement or worsening was small (on the order of 0,QOOOlO inch
or 6.254 pm): '_ .
Bffect of Burr Size on Edge Radius
,,
In the third study, simulated burrs were used to determine how
burr .height and thickness affect the deburring psoc-ess. Cylinders
0.375 inch (9.525 mm) in length of ,303 Se stainless steel were
used. Discs punched from O.OOl- to 0.00%inch-thick (25.4 to .'
‘!
a
J 9
MRR77-12 :
49. -48-
127.0 pm) 302 stainless steel were spat welded to the-end of each
cylinder to simulate a burr, All discs were 0.375 inch ia diameter
-. and the Cylinders were ground to diameters which provided from
O.OOI.--.to 0,010-inch-high (25,4 to 254-O /urn>~tiulattad burrs.
c Height~hf tbe%e simulated burrs was measured optically by rstatlng
the part to find the highest proJectI.on fsnm the su~‘ruc~. &3~:al13e
the discs were not 'accusa-tely center~~d, some runout rsxisted which
made it necessary to measure the highest projection. Alli measure-
ments were made to the nearest 0.0001 inch. Burr height W&S,
measured after each 5-minute Marperizes cycle. The parta were
subjected to the N14 asZumi.numoxide nuggets and the IA-1 camprjrlnd
at 1s g"S. As In all studies recorded in this report, the
specimens and media were covered with wa'ker.
Although the initkl:burr height varied consldsrably, the rate of!
reduction in burr height depended upon 'the -thickness and ccxvtour.
Consequently, the deqta was regrouped intw like thickness and
converted into change in height at time Vi' minutes, using the
equation
Aht = ho - ht
where
Aht - change in height,
% = initial buFr height, and
ht = burr height a$ "t" minutes,
Nine specimens of CLooi- or O.OO%inch-thick burrs were deburred
and a radius generated at the edge. This was reported as a
radius instead of hkight, Theqe values ,were converted into.a
negative height. :,
ji The ,high @;Lnt on the burr i&a ,point of symmetry (Figure 27)'.
which is also used as the referaacs point. When the burr is
remmed, a radius is generated %~eultic&is cotiverted to a,negative
height 'in,t~nps of R. ir.
A. h = -0.176 R " _ ',' (14)
4 ,.
he.following mode3.:.+4:'used to Tit th" data: ,,,'', :i ._ ~t
(15) '; ',T
4:
',
50. --49-
w = width sf burr in miPs,
b = constant, and
This model. adjusts for burr width; a burr twice as thick as
another does not take twice the time for th@ same reduction in
.height. Also, thin burrs have more time effect than thick burrs;
therefore, the power of time 9" includes the burr width "WI' to a
pQW8r.
The resulting equation was obtained omitting the reported values
for"t = 5 minutes (the inspector recorded average height rather _I
than maximum height at this time interval).
Ati=h -h t = 4.69w -1.2
0
t0*?13wO*16g x IO-4 ia&
. (16) -
51. -SO-
The standard error of -t;he estimate was aAh - 0:000195~Lach
(omitting t = 5 minutes). Note that w must 4~ expressed in ~CXII-IS
f of O.'OOl inch units (i.e., OS00175 = 1.75).
&X-t3 giV@ll il.l IhGh WiitS,
The terms b and ho
For example, what is the est.imatcd burr hittQ-ht after l-2 rrrfnut;es
af harperizing under the GonditkxIs of this test i.f the I.nitial
burr‘height ho = 0.0027 inch, and In.it:Lal burf wridth o:r thicknc~:~u
(w) =t 0.00175 inch?
Under these conditions, Equation*16 became5
h= 27 x 1O-4 - 2.396t00784 x 10-4 .> .
andPat t = 12 minutes,
h= 27 x 10n4 - 16.8 x 10-4
= 10,2 x 1oe4 inches burr height after 12 minut;es Barperizipg.
What, then, would be the estimated time to remove %he burr, that
is, to obtain h = 0 or Ah = ho?
By rearrangJLng Equation i6, an expression for the tbn8 (t) require$
to reach a burr height of ht can be obtained: .'
54. Finally,
t = exp
Harperizing has been used to deburr over 150 cliJ~.ifc~ri~kll; COm~Cjni;nl;:-;
at Bendix. The materials represented by these cmmponents include
brass, 300 series stainless steels, 17-4 PB stainless steel ~
Kovar, Paliney., As SCC?Il
in Figure 29,
beryllium-copper, albminuqj, and ceramics.
these parts have a wide varaety of shapes, The
production observations support the dimensional changes described
on preceding pages.
Consider the parts shown in Figure 30, These *parts are shown
next to a United States dime. A IS-minute HarperizE3 cycle in
l/4 inch (6.34 mm>plastic pyramids, N24 aluminum,oxide nuggets,
and IA-1 compound at 15 g's removed 0.0003 inch (7.62 pm> stock
from one dimension which has a tolerance of only 0.0004 inch
(1.02 Gm>. It is obvitius that such a part must'be intentiorrally
machined oversize,,.if print dimensions are to be maintained. This
part is made from 17-4 PH (H900) stainless stee&:"
~~,,4/>r$j,?,?;N
A 20-minute cycle in Nl,d nuggets at 30 g's rembv%%!I0.00008 inch
(2.0 pm) from the 0.025 inch (635 pm) 17-4 PH stainless steel
journals shown in Figure 31. :,
A change of 0.0005 inoh ('12.7 pm) occurred on the diameter of the
fine edge blanked aluminum spacer shown in Figure 32. These parts ' I
were deburred in 20 minutes at 15 g's, using 3/16 triangles and
lLA-1 compound. The: burrs on this part were 0.0035 inch "(88.9 pm>
thick at their root &nd 0.008 inch (20.3 pm) high; the final edge
radius .was 0.010 inch (0.254 mm), m,
Although holes smaller than, 0,040 inch (1 mm) do not usually deburr
completely, those shown in Figure 33 did. This powder metal
ferrite core is soft enough that abrasive particles can *effectively
.?
. . ,
‘5
MRR77-12 -I
._ ’ ; ,, ._
$
1 .,’ I
,,.
55. Figure 29. Production Parts That Have Been Harperized
produce a small radius on the four 0.020-inch-diameter (0.51 mm)
holes. A stock loss of 0,003 inch (76.2 pm) occurs in the center
hole using N14 nuggets and lA-1 at 15 g's for 20 minutes. A
a-hour run in 3/16-inch ceramic bonded triangles produced a
0.020 inch (0.5 mm) radius ,on ‘&'3Ainch-diameter by.O.S-inch-thick
(76.2 by X2.7 mm) ring.
Effects of Piece Pa& Geometry
'. Asia all loose abrasive processes, deburring aggressiveness is
a function of part size and media size and shape.
if slots and steps are too shallow,
As an example,
,occur (Figure 34).
complete deburring may not
In this case the step height TItl must be
greater than the media radius RS -for 'deburring to be effective.
If triangular shaped media is used, the slot width mu&exceed
,
MRR7f42 '
56. ‘
Figure.30. Miniature Stainless Steel Blocks-Before.
and 'After Harperizing L J "
c .-i‘J*t ?RS. (Actually, the slot width should exceed 6 RS to.,prevent
media lodging in holes.) If these features were the bottom of a
counterbore, no deburring action would bocur because the media r
cannot readily move out of blind pockets.
*
While reducing media size may seem an obvious solution, the
deburring forces decrease rapidly with media size. For most
purposes an Nl4 nugget is as small as one can easily-:use (roughly
+a particle of 0.060 inch (1.51 mm)]. I I
a 1
57. Figure 31. Stainless Steel Pinions Before and After Harperizing
As reported elsewhere3 radiusing is a function of the angle
between the surfaces (Figurg 35). Provided the part does not
shield the media an edge angle of 120 degrees can develop a
radius 10 to 20 times larger than an edge of 30 de‘krees (Figure 36).
While the Urge angle develops a radius.faster, it is also much
more difficult to control its repeatability.
.Large parts tend to bang tqgether in small barrels. For this'
c reason it is not advisable to deburr parts Larger than 1 inch
(25.4 mm) *at speeds-above 5 g's in the machine used in this
study.
.
Long ) thin, sdft parts also tend to become distorted in the
Harperizer. In one'instance, some 3-inch-long (76.2 mm) wires
were actually tied in knots after Harperizing. While lower
speeds reduce the tendency for damage, the potential is always
there for large parts .in -small barrels. The initial jerk as the
machine begins tQ rotate i's kuffitiient to dislodge shock and
1a
MRR7742
,?‘:
Y
I
i
‘
,
58. FiTure 32.b Aluminum Spacers Before and After Harperizing
acceleration indicators (Impact-O-Graph Shock Monitor, Impact-@-Graph
Division, Torq. Engineered Products, Inc., Bedford, Ohio) set for
30 g's even though the machihe is set for 5 g's.
One of the major limitations of the process for use on precision
miniature short-run parts is that it may take 2 to 4 hours to
separate nonmagnetic parts from media of roughly the same sizrj.
While this is a problem for only a few shapes and sizes, it can
be a limiting factor to usefulness.
Barperizing Effects on Subsequent Prbcesses
A'11 loose abrasive processes'beat minute.particles of abrasive
media into part surfaces. While these:.are not visible under 200X I
magnification, they can be detected by Auger Electron Spectrometers.
This contamination which is not removed by conventional cleaning"
processes, can cause blow holes or stress concentrations in -1
welded joints. It can also prevent good.braze joints (aluminum
oxide, which is,the most commonly used abrasive, is often inten-
tionally used on tooling to prevent braze wetting on tool surfaces). -
Recent studies, indicate that the prsbability of adhesfon failures
,increases when 'difficult-to-platel!lmaterials have such codtamination*
1 MRR77-12 -
L *'. 5
59. -o-meTiICK~
(0.51d,..,
Figure 33. Powder, Metal Part
f
I”
(4.57 Itmj
r 0.045 INCH
(1.14 mm)
on them. The barrels used in Harperizing have enough aluminum
oxide abrasive in their surfaces to contaminate several lots of
parts. For this reason it' is necessary to have two sets of
barrels so that one set is onl,yj used for non-aluminum oxide
abrasives.
While aluminum oxide media is the worst offender
b
silicon carb'ide
and quartz base materials also cause ome problehs but on a
smaller scale. Dolmite (magnesium calcium carbonate) is the
only 'material studied which could be cleaned from parts.entirely
* without chemically,etching p-arts. Plastic media left a thin film
of plastic on the parts studied.
Since most parts are not plated or welded, aluminum oxide media
does not present a problem. Because it lasts 10 to 100 times
longer than other media, it is generally the most economical.
Silicon carbide and dolamite break down quickly into small pieces
which lodge in small holes, .grooves, and similar crevices, No '
media appears to create adhesion problems for dry film lubricants
when they are applied after 'Harperizing.
a
L d
I
MRRl7-12
60. ABRASIVE
WORKPiECEi= MEDIA mDURRm
WORKPI EC&= AORASI VE e;r
t =(4IEF’THqlF GROOVE
b = WIDTH LoF GROOVE
ABRASIVE
CHIP
a > ry = OEBURRING UNCERTAIN
Bl? B?, ‘83 = POINT OF CONTACT
.1 ’ ABRASIVE CHIP-HOLE RADIUS
RMIN .* MINJMUM HOLE RADIUS
WORKPIECE.=; MEDIA= dURli~
RW
= RADIUS AT WORKPIECE
- .- .___.
RS = RADIUS AT ABRASIVE MEDiA
Figure 34. Effect of Part Geometry bn Deburring
‘'a
.
MRR77-12 *- ",.
v'_',.
/
61. -6O- .
.‘I /---- STOCK REMOVAL REQUIRED TO GIVE INDlCAfEb RADIUS (S)
RADI US
‘,
I
INCH
---------
0.002
0.003
0.00s
O”O10,
0.015
6s AMOUNT OF,MATERIA!. REMC)VEU
’
’
,
/
50.8
76ea
127.o
R54.0
381 .o
0.0008I 20.3
0.0012 30.5
0.0021 53.3
0.004P104.1
0.0062 157.5
0.002
0.003
0.005
0.010,
'0.015 /
‘ED
I:b:I2'7l Q 0. 0
'f&4*.0. 0.0387 .983.c
.-3U.o' 0.0581
I
1474.7
62. O.Ocll
0
.
50.8'
25.4
0
0 1 2 3 4 5
TIME IN VIBRATORY FINISHER (HOURS) .’
I.” ;
Figure 36. Effect of Edge Angle a'nd VfbraTion' Time on Edge‘ .
Radiusing of Phosphor Bronze Workpiece
Harperizing has several positive advantages that are often over-
looked.' It removes heat treat scale or tints that may interefere
with appearance or passivation. It removes machining oils and
die lubricants (the abrasive oompound contains a large proportion
of'soap). .It can be used to "machine" oversize parts to a
smaller size. It imparts residual compressive stresses which
, 'generally improve fatigue life. 'In some cases th$s can also
result in part dist-o&ion. It can produce finishes as fine as
2‘microinch AA (0.05 pm). In general it improves finish& while
cieburring but further improvements require special burnishing
cycle?. .
Whilefine sur$ace finishes may not be necessary one user'has
pointed ,out that they do have two beneficial side effects. Every
individual who handles the parts afterwards associates fine f. ,*
MRR77-12 I -
--
i
63. -62-
finishes with precision and hiigh quality. This in turn causes
most individuals to treat such‘parts with more care. Secondly - '
I parts'with fine finishes are easier to &ean since fewer crevicr+s
and scratches exist on them to trap m~.l.,s arrci c:nntm.inants"
c Evaluation of Published Literature'I_l-- ---."--
ACCG~ElSBMENTS
The effect of hentrifugal barrel finishing par~et~:rs on ~urk'acs
finish, edge radius, and dimensional changes'has been determined.
The influence of burr size on the time required for deburring has
been-noted, and typical production limita%~ions and applications
have been described. Culdelines for the implementation of this
process have also been prepared (Appendix,A).
FUTUREWORK
Although no future work is planned as part bf this process
development progect, some additional'documentation of,process
effects on size changes. from this process will be made. Beciu,se
.of the increasing need ta produce surface' finishes of I6 microinches .
(0.406 /ml), a future study would be desirable on the centrifugal ', '
barrel finishing parameters required to produce such finishes. I
c
5
‘.
64. REFERENCES-_I.
.-a 'M. Matsunaga, "Theory and &xperziments on Centrifugal Barrel
Finishing," 1ntsrnational JournaZ of Production Reoenrch., Vale, Fii
No. 4, 1967, pp 275-287,
Q’rQ
'6M. Matsunaga and Y. Hagiirda, Researches on Barrsl ,"ini.ohing,
Report al The Institute of Industrial Science, The University 0%
Tokyo, Vo3. 17, No. 4 (Serial No. ill), February, 1967, pp 145-154.
'F, Schafer, Product Design Influences on Deburring, SMl3 Paper
‘. MR 75-383, 1975.
*G. B. Lur'e and A. P. Sinotin, "Tumbling of Workpieces in Drums
with a Planetary Rotation," Russiun Engineering JournuZ, Vol. 54, v
No. 8, 1974, pp 39-51.
'5. Bernard Hignett, Cap,abiZities and Limitations of Centrifugal
Barre2 Finishing, SMEPaper MR 75-834; 1975.
!*H. Schmalz, "Centrifugal Grinding and Finishing," GaZvanutohnik, '
VOL. 63, No, 4, 1972, pp 325-34 (in German).
I
.
5
1
:. MRR77:12 I ‘-’
QrI
‘. .,.
65. 106S1380 A1.$13 triangles, 3/a by S/15 by l/4 inch ('3.5 by g.ti
.by 6.35 mm), AX90 composition
10651569 A1203 triangles, 5/8 by 5/8 by 3/16 inch (15.9 by 15.9
by 4.8 mm), AX90 compositidn uf?!
1065136A A1203 fj8-inch-diameter by 12/32-inch-long (3.2 by
8.7 mm) cylinders, Almco Cl-8
10651383 Vitrified A1203 angle-cut cylinders, l/$-inch diameter "
by S/B-inch-long by BO", AX90 composition
EO6El.387 Silicon carbide triangles, 5/8 by 5/8 by 5/16 inch I
(la.9 by X5.9 by 7.9 mm), Fortune Industries CS-46
10651435 AJ.20 angle-cut cylinders, 3/16-inch diameter by
1,1/3?!-inch ldng'by 45', AX90 composition
10651605 Fused,Al203 chip, Carborundum number 6, 0,092 to
0.188'inch.(20.337 to 4.775 mm)
10651566 Fused A1203 chip, Carborundum number 7, 0,078 to
0.156 inch (1.961 to 3.962 mm)
10651568 Fused Al203 chip, Carborundum 8 grit', '0.062 to
0.130 inch (1.575 to 3.302 mm)
10651575 Fused Al203 chip, Carborundum 12 grit, 0.045 to
0.090 inch (1.143 to 2.286 mm)
10651560 aAl2O3 chip, 14 'grit, 0.040 to 0.060 inchm(l1016 to
1.524 mm), i'vlechanical Finishing Company liuumber
;281(B)-14
10651,561 Fused Al& chip, 24 grit, 0.020 to 0.030 inch (0.508 " ,
to 0.762 mm), lliiechanical Finish'ing Company Number :
374(R)-24
10651260 .Silicon barbide chip, number 8 grit, 0.062 to
0,130 inch (1.575 to 3.302 mm)
.a
MRR77~12 , '
68. 8.
Table A-3. Selecting Centrifugal Barrel Finishing Conditions
continue
Edge radius xtllowable
Dimensional change a9lowable firom this proce$s ._.--.-
Surface finish requirement (See Sectj+on AZ>
Continue
4 What is the initial. part qondition?
Burr thickness
Burr height
- Surface finish on critical surfaces (See Section A3 and
Table A-6) ,.
continue
5 Which deburri.ng parameters will produce the requirements of
Step 3 with the conditions listed in Step 4?
(See Section A4 and Table A-7)
.”
s
6 Is the part subsequently welded,.plated, or used as, an
electrical con'tact?
.*
NO ‘Yes - See Section A5 1 I
c
7 Are the selected. conditions adequate in practice?
Yes, Xo - See Section A6
MRR77-12 *
69. Table A-3 Continued. Selectixlg Centrifugal Barrel Finishing
Conditions
The random shape m&dJa con,sists of media of many dif-
ferent sizes. Even though it is graded by size, a
wide variation in size and shape exists (Figure A-2).
Even the preformed shapes such<~s triangles and
cylinders have some variation &I size.
As & general rule, select media which is smaller than
the part. Ideally when holes or slots are present
media should be either
9 Slightly larger than hole size, Q
a Slightly smaller than hole size, or
‘.
+ One-third the diameter of the hole.-%
Be&use 'of media wear, the third choice is the safest
with random shaped media. For preformed shapes, the
first two choices are easiest to assure.
For general usage on precision mi%iature parts, the ,.E"'
N14 random shaped nuggets are the most desir ble size.'
If hole size precludes th& use of any of the media
shown in -Table A-2:
i Screen the media to remove oversize or undersize
particles (Table A-4); t ,
Yh
a Plug holes with rubber plugs,‘rubber tubing which
expands when tension is released, rubber bands, or
.nylon line with ends expanded by heat (Not& that
some media in holes is not harmful if it can be
_' 'easily removed.); a
'. -
c NRRT7-12
70. -69-
A2
Allowable dimensional ,changes of at least 0,0002 inch
(5.1 Drn) are desirable in this '$rocess. As a lower
J .& imit , it ,is possible to maintain sitie within O,OOOO5
inch (1.27 pm> if burrs are small.
ALlowable surface finishes should be 32 microinch'
1 (0.81 pm) or rougher, although it appears possible that
finer finishes can be produced or maintained.
Corner radii (therintersection of three or more sur-
faces) (Figure A-3) will be 3 to 5 times larger than
edge radii, '.
Vhen'a tiorkp%ece has edges with several different
requiremerits, the centrifugal barrel process‘ should be
vased on the most tightly toleranced features.
I
Note: Internal features will not receive as much
action as external features.
A3
s .
Initial Part Candi$k&
.
. Centrifuial barrel finishing cati roughen kurfaces which
.were initially 8 or 16 microinch %(0.210or 0.41 pm).
Deb&ring knditions are *obv'iously depindent on the
size of thb burr$to be removed.
,
'If burr'size is.not 'known tde value's used in Table A-6
will provide some guidance. I?
, .
a
VViRR?f.-12
q l
L
I.
71. ,r
Table A-3 Continued. Selecting Ce ifugal Barrel Finishing
Conditions ,.'
i bt
A4 Burr size, allowsble stock .loss # and muximlm a1Xowab.lB
edge radii are the msst significant Pac,tors in datf2r-
mining which combination o.f puw~~~ters will ~W~UCO
acceptable results. Burr location and workpiece
,geometry are a;bso important, but at this ,time data jt,
._ not avallabl.e 'Co indicate their quantitative signlfl.-e
cance.
Recommendedcentrifuga~~ barrel finishing ~ara.met~~g ax+4
shower in Ta'ble A-7. These secnmmendatinns are based on
the data,shown in Figures 4 through 26, the study of
effects of burr size, and produclion,experience. hs
seen. in Table A-7, it ii not possiblr; to remove burrs
0.005 by 0;005 inch (I.27 x 127 pm> while maintaining
.a 0.00005 inch (1.27 pm) ma.~imu.rr++,~stockloss.
These recommendations indicate the most aggresskve
combinations which can be used, In some eases it may
be possible to lower the time, g level, or media size
and still produce the desired results.. With the
exception,of Large allowable stock losses, conditions
indicated should be the condition tried. If the
recommendatidns are not quite sufficient, changes
should be made using the' list of problems and solutions
given in Sect~ion A6.
A5 Aluminum ,oxfde media contributes to blow holes ,and
stress concentratLons in welded joints. As such it
i&,desirable to use other materials when such parts
have to be deburred. Silicon carbide or dolamite,are
acceptable alternate materials.
BGaziag~ and soldering are also affected by minute
aluminum Ioxide particles.. Solder will.not adhere to
" tt surface having such particles in or on it.
. :.
Theaprobability of plating adhesion failures increases
l
* Pwhen,alu+num oxide is present on surfacep., This may
il not be a significant problem on copper b$s'e.alloys
since they activate .easily. rT.tiIs.,,&an .be a significant ,'
I ' problem on any hard to activateu'm&t'&~al,
i
The adhesion-of solid film lubricants and the passi-
-
c
vation of.stainl,,ess steels do not appear to be
ected,by the 'presence of minute aluminum'oxide ;"
particles.
0 MRR77'=W2,: *
‘9
72. Table A-3 Cent inued . Selecting CentriPugal Barrel Finishing
1 Conditiuns
* - _~.-“e---~-~-.--~ .m_I-c”,_I------
Sect ionv DiSCUSSi.0~
--..“_ ,.“._I.. ..- ---- - ._,...-.,,x “_ -.---..._- .- -l..-.-I. , :__
Use larger media;*/
Run for a longer time;
_I qi
Increase "g" level; k$
Use a combination of 1arg.e a,nd small media;
M&nualLy remove large initial burrs;
Use a more,abrasive compound;
Be sure the media shape is consistent with part
geometry;
Use.a lighter media i$ burrs are bent over;
Place the parts in a fixfuye; or
Rkdu,ce the water level.
solve surface damage, impingement', or rouihness:
.I
Use smaller pedjj.a or well-worth &edia;
w
ed Reduce the cycle time;
.-
l
a
*
When usiqg two different sizes of media,.use more
of the smaller size;‘-..,
,
Use a less abrasive compound;
3elect a more appropriate media shape; -
MRR77-12 .' "
. *
n
i,
73. Table A-3 Continued. Selecting Cantsifugal BsrreL.Fi.tiishing
CaaditI.tonst
8 Mask sensitive nrtjuf;;
m Increase the
8 Reduce the n
j To keep the media
8 Use sizes of
and recesses
* Do not‘use U?regularly-shaped media where lodgi.ng
is a problem;
8 Soreen the media psiai to @se if lodging occurs;
and bb '
8,‘ Choose a medi,a"s,ize which can be screened or
separated from the parts.
To prevent burnishipg:
8 Do not use steel burnishing shot to remove heavy
burrs; and
8 Do not mix'steel burnishing stock with abrasive
media. b
To'keep flat workpieces from sticking together:
a Use small random shape media.
I$esidue ..,a ,,
.
.
Occasionally parts may have a white residue left in
recessed areas. This is generally some of the abrasive
or burnishing compound which did not dissolve in
tirater: It can be removed by a thorough ultrasonic
oleaning inhot soapy'.water. ,' It can be prevented by
using liquid compounds in the burnishing cycle. "Liquid
compounds are generally not as effective as the
, powdered compounds, however,, for burr removal or
burnishing.
I
74. Table A-3 Continued. Selecting Centrifugal Barrel Finishing
i Conditions
Distortion
Long flexible burrs produced as$;;acutter passes over
the back edge may qpt be removed from some parts.
These burrs tend to double aver and become U-shaped.
This now-reinforced burr may be hammered flush with
one side, but will not come off. These burrs, which
occur frequently in the soft metals, should be removed
by some other process. Beating them flush will only
mak& them harder to remove. The use of plastic media
rather than the heavier abrasives will minimize burr
double-over,
MRR77-12
80. Table A-6. Burr Sizes Typically Found on Precision Miniature
Parts at Bendix
Screw Machine 0 *001
Turning (hand feed)* 0 *0045
End ;Milling** O.OU35
Side Milling** Conventi.onnl. 0 *0015
"Master Cut" Cutter 0. ool.5
Drilling** 0.0030
Remning 0 . 0005
Grinding o.ool.5
Ballizing 0.0010
17-4 PH (H900) Stainless Steel (R&2)
..,
--Screw Machine 0.0010~ &XL4 0.0015 38.1
Tur.Lng (ha.nd feed) cl. 0017 x3.2 0.0018 45.7
End Milling 0.0020 50.8 , 0.0100 254.0
"Master Cut" Cutter Q.0005 12.7 0.0002 5.1
Grinding 0. oois 138.1 0.0020 50,8
GOBI-T6 Aluminum (Rg90)
End Milling.
Drilling
Reaming
*0.0030 76.2 0.0060 152.4
0.0030 76.2 0.0050 127.0
0.0005 12.7 0 ..0030 76.2
1018 Steel (Rc20)
'End Milling 0 10035 88.9 Q.OQ80 203.2
Drillzing 0,003g 76.2 0IO055 139.7
Grinding 0.0040 101.6 0.0020’ 50.8
*If lathe operators exercise good machining practice they will
;remove the majority of this burr, by wiping sandpaper or simi- '
lar abrasives over edges before the part iszemoved from the
lathe. Data for -screw machines is .an estimate based on e"sti- *
mates' from production experience. All other data is based on
over 20,000 measurements.
**At the end of a cut a chip typicaLly rolls out of the path of
'4
the cutter and produces' a long roll-over burr, This burr is
, not ,,included in the values shown. In the case of milling up
to eight different burrs are produced with a single-cut--the '
values shown are reasonable worst case values. '
a
__ ‘-
I
: MRR7i-42
= I
82. ,_.
-81-
_1 ,
,.
MRR77-12
+
Th! 1Society shall not be responsible for statements or
/
wmons advanced in papers or in discussions at meetings of
the Society or of its Councils or Chapters or printed in its
publications.
‘.
..,....*.
