K. Hitomi
one
LOST Class fivagen pos
Reliability Analysis of Cutting Tools
‘This paper is concerned with the reliability analysis of tol life based on the tool-wear
tales obtained from metal cutting experiments. Fram experimental results, statistical
distribution of tool wear was decitied, and the distribution of toot life and the reliability
Function of cutting tools were derived. Further, t was shown thatthe reliability of cutting
tools ai @ certain time wos easily caleulated from machining parameters and tool-wear
limite by the use of reliability function
N. Nakamura
Deparment of nda Engen,
‘Semiormacht Histon sh apa
$. Inoue
atcmesi, Onakats, Sapam
Introduction
In mat machining operation, too ile is considered to he a most
important factor. Especially, all the mathematical models of the
rachining operation rely toa greater or lesser extent, on the tol life
‘quation predetermined by experimental methods or others. But,
aside fom the problem of experimentally determining the parametors
inthe too! te equation, there isthe difticity thot there ae then
herent variations in tool life fora certain set of machining wondi
ons.
Since the early and pioneering experiments by F. W. Taylor (1),
there have been several tool-life studies, Wo [2} applies “response
surface methodology” to tol-lie testing for predicting the tol-life
‘curacy for the equited confidence levels. Wager and Berash [3]
reported on tests of hundreds of high-speed stel turning tools and
shoved that tool life yalues are subject to a statistical dsteibution
‘hich can be approximated by the normal distribution. Dosle [4]
examined tool-fife variation inthe ene of high-speed stel taps and
{rod the variation to lek of eanrol in processing, using probability
theory, Friedman and Zlatn (3] reported on the alysis uf to! life
‘aration for several metals and eating tool sworkepice combinations
[etishi and Aoki [6] investizatad the inuence of Feo eate on the ool
‘Centributed by the Pradution Bnioasing Divicom fr presentation tthe
Winter nia Meeting, San Franco, Calf, eeembu 10-10, 878 of TH.
ASME heudguerrs July 24,1878. Papor No. 78.WA/PROD.
Journal of Engineering for Industry
life during interrupted cutting in order to determine quantitatively
the reliability of earbide cutting tools. More recently, Ramalingam
and Watson [7] reported on the single: and multiple-injury tool-ile
‘nodels hased on the hazard function theory
‘Although the above studies made a significant progress in under-
standing the specific nature of tao fe, it sti remains some difficult
problems, Particularly, for precisely predicting the tool-lfe values,
‘nore research on the reliability of useful life of etting tools from the
‘standpoint of probability theory is needed because tool if and tool
‘wear have an inherent stochastic nature.
‘The purpose ofthis paperis to conduct reliability analysis of eutting
‘tools based on the tool-wear vales obtained from metal-cutting ex
periments. From experimental results, the tool-wear distribution and
tho estimates of tool ie are decided, andthe tool-ife distribution and
the relisbilty finction of cutting tocls are derived. Further, itisshoam
that reliability of cutting tools at arbitrary tine is easily ealeulated
from machining parameters end tool-wear limits by the use of rli-
ability function derived,
2 Experiments on Tool Wear
In this soetion, stochastic nature of tool wear is experimentally
Investigated and the suitable type of distribution of tol wear is ts
tistically determined,
21 Bxperiments, The following two types of experiments were
done for investigating the tool-wear distribution,
Experiment 1. Generation of data for x2-goodness-of-fit vest as
‘othe tool-wear distribution,
Bxperiment 2. Generation of data for the multiple regression
MAY 1979, VOL. 101 / 185Table 1 Catling eonations w
Cutting tontitions | caperinant 1 [ experiment 2 oo
Speed (nna) 175 wo
Feed Rate (anirer) 0.20 a.
Depth of Cut tna) 10 18 ané_ 2.00,
Tieasurang Ties oF Te FAG. 10s V5 20, 85s
ian dear 28,035.40) 4B, 50, 5560 s
analysis for determining the tool-wear distribution, ol o
‘The experimental conditions were as follows. cr rr Te
‘Machine took High-speed Lathe (Okuma Co, Td) Cin tae i
Material carbon stee-billet (500 mm lang, 100 mm in di.)
Cutting ton: carbide, throwaway type; Geometry: ~5",
ers 0 7
Cutting mato Dry |
Citing endo See Table wf |
o7
es . _|
Bosh LL ser
_ es 38 aan) aD a
: este te
fa. 2 aor tnbear aes
ee ae
0 Frity experiments were conducted for Haperiment 1, and the
tostmaa cures hes stained ore shown in Pig Ite om the
fourth ol wor crs intersect ah ther bene, he ineresing
orig rae ol ear ies saa anata
“The Maga the ol neat valuenat the meso Gand 45min
areshowm in Pig The looowet vais ze observed tobe ite
{Mea ina marco tie ang a he aly stage, and in ater wide ine
. o 0 40 o = 70 range at the long elapse of time, However, as seen from Fig. 3, the
tong teh) iit of atten fe cnsderea to Beamon constant.
asin faye muna, tn Enso} i EE to esate the fo wear thu the zed
Fig. Tookwoor curves
obit test wes pertormed, considering the normal distebution, the
‘Weibull distibition, and the logarithmic normal distribution, The
Nome ne
sb, bg, Ba, C= estimates of 8, Ba, Bas
‘snd Cy
Gq constants Cu
2 dopth of eut,mm
FL) = probability Fonction of tool wear
{1) = probability density funetion of tool
K, = confidence limit ata
@(e) = probability function of too! life
(2) = probability density function of tool
lite
‘act) = ettaby Sunction
Bit) = derivative of R(¢) in regard tot
1 food rate, mm/ee
486 / VOL. 101, MAY 1979
Ty = mean time of fa
1, = estimates of to! life
Ts = lower and bi
‘of tol life
= euting time, min
Var
Vp = tool ear, mn
q = estimate of tool
Vat
16 = machining speed,
bye critical region
Bud Bes Bas Co
™ variance of lifelength
Timit of tool wear, um
1 in = independent variables
constants in regression
lure eate «= error, = 018)
16) = failure rate of cutting tool
= mean of normal distribution function
018) = error
standard deviation of normal distribution
jgher confidence limits
VotlT
PriVa = VatITol
were is one side riicl ropion
‘Since the tool wen distribution i given by equation (8), the above
uations are transformed as follows:
Law
a
erlnve-mesbanear MOE = a
(seni idVq* a (16a)
STE"
eoubaornnind vn
tas
Using he standard nor dst
In Var = incor 4
»(eueane ) (te
tn Var = dcr
> (varie area am)
where
488 / VOL, 101, MAY 1979
nto fi ew
Suppose that 8()
(17) aze expressed
" = [Wp (CabroPraibe=lebo
Ty = [Var CoPohasyibneoan
1 ey forennstant Ky equations (170) and
(18a)
(186)
Hence, rom equations (12), (282), and (185) the contdencemnterval
‘of the tol life i obtained fr estimation of ite of cutting wok:
Cott s TS Cstt a9)
where
Cas elie
1 should be noted that the confidence interval is not determined
by Ta whichis obtained from the expected value of tool wear, but by
TT, which is obtained from the generalized Taylor toolife equa:
4 Reliability of Cutting Tools
Tin this section, the reliability of cutting tools is analyzed, using the
tool-wear distribution, This is the probability that a cutting tol
Dorforms its furetion adequately forthe peviod of time intended under
the operating conditions encountered.
“Li Derivating the tool-life distribution. The too-ie dstri-
tnuton is desved as fellows fron the tool-wear distribution, which was
determined in the preceding sections.
Lethe peobebility of tool is and itaclonsity Fametion be denoted
by Git) and g(t, respectively. The end of tol lf failure is judged.
Dy the limit of tool wear (Vo")
‘Suppoce that a cutting tol begins to function atthe time period
‘of 0, and that its failure yecurs at 1. Then, the probability that the
futting tool wears out prior tothe time period of, i given by:
Pri
«|
Pin tay
@
Using the above Function, the reliability ofthe cutting tool i as
Follows (9
1 ‘The mean time of failur-feee operation:
So ane
sn (28)
Rind
9 Varianen of ielength:
Journal of Engineering for Industry
pa
Bas L
wo catia Tine (ain)
(Spon 175 fai, Fod Rates 8.20 aire, pth OF Cu: 1.50)
(Lint of 1003 corm)
ig. Comparison of the experimental andthe calculated rests for vel
ity and ators rae of exting toa!
Var [|= Bt ~ Tal =f s89(eiae
=2 i tRUdt — Ty?
ery ws
4 tamaninatnccrioia
oe
No
wy ae a
aa RO)
43 Experimental analysis for reliability of cutting tools,
Based on the data for tool wear shown in Section 2, se compare the
characteristics of reliability derived theoretically from the tol-ite
slistribution and those obtained experimentally from the tool-wear
distribution
Experimental and theoretical values ofthe mean, variance, cosf-
ficient of variation, reliability, and failure eate ofthe cutting tol nee
‘shosen in Pabie and Pige§ and 6
From these result, the following conclusions are derived
1 Experimental and theoretical values ofthe mean, variance, and