You are on page 1of 7

Engineering Costs and Production Economics, 2 1 ( 199 I ) 5l-57 51

Elsevier

Strategies in spare parts management using a reliability


engineering approach

Anwar K. She&h
King Fahd University ofpetroleum and Minerals. Dhahran 31261. Saudi Arabia

Frank L. Callom and Salim G. Mustafa


Saudi Aramco, Dhahran, Saudi Arabia

(Received January 25, 1990; accepted in revised form July 5, 1990)

Abstract

Maintenance in a process industry, such as chemical and oil companies and food processing, needs frequent part replace-
ments to assure continued operations. These parts represent non-repairable items which are best purchased, such as bear-
ings and gaskets. Based on limited information on the time between failures of such systems, one can forecast the need for
spare parts for a given planning horizon, which may be from one to several years. In this paper we discuss the methodology
to calculate the spare part requirements for non-repairable systems and purchasing strategies for these parts. These strate-
gies are based on a stochastic characterization of the time between failures of the system. The inventory level to be main-
tained will depend upon the need of such parts in consecutive time intervals or inventory cycles to satisfy the maintenance
requirements with an acceptable risk level.

1. Part reliability and failure analysis riod, the conditional probability of a surviving
unit in an additional time increment, At, remains
Reliability of a part is the probability of sur- constant. Thus there is no wear out or aging pat-
vival as a function of operating time t. Typical tern in the failure data. A new component of this
reliability curves for various industrial products type put in operation at time t, will have the same
are shown in Fig. 1. In service, the components chance of failure in time interval t, to t, +At, as
fail randomly which is reflected by the failure an old unfailed component at time t,, which was
rate. Often the failure rate remains constant with put in operation at t =O. The most appropriate
respect to time. That means that for any time pe- replacement strategy for these types of parts is to
Inn
replace them upon failure. Some of the compo-
nents which aproximately follow this type of fail-
ure pattern are:
6 Fan BeltsFor Ca
1. Electronic components
2. Control devices
3. Ball bearings
4. Light bulbs
5. Shaft seals

2. Determining the failure rate and failure


intensity

Fig. I. Reliability (probability of survival) for various indus- Once identification of the parts with the
trial products. (R(t) = reliability; t = life; r= average life) equipment is made, strategies can be formulated

0167-188X/91 /$03.50 0 1991-Elsevier Science Publishers B.V.


-suba aq$ ‘sa3nl~i?jI.KWM~~q awry Jo ~aporu [eyiau
L9’0 a%eyea~9
-odxa ue ~03 snyL wed a~130 aw.~ amIFc3 aql sy
EE’O J'?aMS
S’O UO!SOJJO~p y alayM CT/I =J pm ‘ I =_y t3upm.j ‘asm pz!Dads
0’1 lua~alll!Jqw3 (PI B s! amI!%? uaak)aq auy3o ~apour p?yuauodx3
0’1 %uyJng (a)
5’0 uo!lepw%aa (4)
EC’0 fww (e)
blgenb Ie!.tapzutJo a%ur?~ E
a3nw3j z
uoyauJoJaa 1 si I ar.uy UI
sa.ytz3 30 zaqumu 30 uo!v+ap p.wpuws ayL
(2)
( [ pl uroy pauw
z 318V.L
01 ampal ~I!M (9) wba ( 1 =x) laporu Iyluauod
(I) (I- Z x)~+~-(1)N13=w’H
-X3 Ut? JOd a 1 alCJI?J_ UIOJJ pt3.I S! (d) , _ @I CLG3~M I 1--
‘ [ z ] se passa.tdxa aq uw uoy
-3UtZZ~“M~U~.4 3Q SE UMOUy OSIpl ( I) “Jj = [ ( 1 )N ] 2
samge3 30 mqrunu a%mam ar.p uay$ ‘pouad sy$
Bugnp apeu.~ aq 01 paau swauxam~dar Imanas pue
Lq ua@ s! ‘d- 1 01 ienba %UOIalfnb sf 1 awfi uogemdo aylj~ ‘(J/ (J )D =x
a%moys JO hI!qeqoJd e ~J!M poflad s!q$ %uynp ‘sa_ml!e3 uaaM%aq auy 30 UO~+IEA 30 iua!3
papaau N sa.nzds JO laqwnu aye uayl ‘uoz~_~oy -rJJaoo) (J )r, St?sa.mlItz3 uawlaq awyl JO uopt?~~
%u!uueld I? sluasa.Idal ( s-1 ) .suba u’f ,I avuy JI -ap p.mpuws pue L st! a.xnIp23uaahlaq atuy a%wa
-p3%UE!y3utI U!W.K?.I IIFM ( 2) *uba pue -AB ut! %u~~vq vJt?;i E 30 s)uawaz?[daI Japtsuo3
mpaJepwIt?Daq ue3 dtg!qEg%? &red amds3o saggI_q
-eqojd ayl ‘33nI3tUlS a%?Stt wed acuos q$t~ luauI
-d;nbaJo uo?l~lndod ay$ IOJ pau!uua%ap uaaq say
ale1 ampe3 ayi aDug wysy~ls “auf.xaau@uavmld
-III .IO e~t?p 8uywymm UIOJJ samI+? palDad
(P) -xa 10 salm a%sn pm!.~o)sg uo past?q )led ayl30
a3g a8e~a.m ayl %U!dOpAap Aq paurwa~ap aq ~23
yyl *swd DgrDads ayl3o am amIte3 ayl ICJguapi
0% St ‘laA3MOY ‘da%sJS.UJ ayL *&god hiO$uaAu! UO
OO’P 66'66 OZ'C E6'66 60X 06'66
00x 98'66 88'2 08'66 SL'Z OL‘66 S9'Z 09'66
LS'Z OS'66 OS'S XC'66 OP'Z 81'66 EE'Z 00'66
OZ'Z 19'86 SO‘Z 00'86 OO'Z ZL'L6 88'1 00'L6
SL'I 00'96 S9'1 00'56 09'1 ZS'P6 9S'l OO’b6
OS’1 ZE'E6 85‘1 00'06 SZ'I PP.68 bO’1 00.28
00’ I EI‘b8 b8’0 00’08 19’0 OO’SL 00’0 00.0s
(a-@ 001 Xd f4-@ ooixd (4,-q, OOI xd 001 Xd
(f E] “0.13) d JO satyen snoyeh JOJ (d) ,_Q JO alqel
I FI78V.L
ZS
53

usage rates. Usage rate of a spare part X, associ-


ated with a specific machine j can be calculated,
NC_=:+ (7)
by using historical information of spare part uti-
lization in the past, with the help of the following
Coefficient of variation K= a( T) / r, can either equation;
be calculated from historical time between fail-
ure data of the part, or in case of absence of time
U(t)=-
cx, (8)
between failure data, initial estimate of K can be
made from [ 41, in which the information about
tcCQ,N
the scatter of life is given in terms of the Weibull where
reliability shape parameter /I. The approximate
relationship between K and p is KZ 1/p. K varies O(t) = Average rate of utilization per in-
from 0 to 1, K= 0 means no scatter in the life of stalled part per calendar month = I?
the parts, whereas K= 1 means a significant IX, = Total net issues for spare parts X, to be
amount of scatter. Mechanical parts usually fail used on machines j
by one or more of the following failure modes, t, = Number of calendar months for which
i.e., deformation, fracture, aging, corrosion, wear transaction history is available
and leakage etc., therefore some representative Q,N = Quantity of machines of type j using
values of the coefficient of variation K corre- the part X, multiplied by the Number
sponding to these failure modes are given in of parts X, per machine
Table 2.
Once the U is established there is a need to link
Example 1 the usage rate based on calendar time t, of the
part usage to the failure intensity of the part which
Consider a system in which simultaneously 10 is based on actual operating time t. In order to
ball bearings are used (K= 1). Failure intensity establish such a link we define the part utiliza-
of these bearings is 200/ 1O6 failures/hr. Calcu- tion p as follows:
late the number of bearings required for a contin-
uous operation of 2 years. Desired confidence (9)
levelisp=l-0.05=0.95 [5].
Solution: Failure intensity of a ball bearing,
where t, can be any convenient life period (often
1/r, = 200 x 1OP6 failure per hour. Average fail-
consider as one year). Since t, > t, the range of p
ure intensity of the system:
will be from 0 to 1 (0 <p< 1) The usage rate
.

l/T=Cl/T1=10 (200~10-~) failureperhour. V(t) = 0, and the average time between failure F
(or average intensity of failure= 1/r) is related
Number of hours in one year:
as follows:
2x365x24=17,520 hr; p=O.95 and Q-‘(p)
= 1.65 from the tables of the normal distribution
(Table 1). o+(p) (10)
Substituting the values in eqn. (7), we obtain
N=35.04+ 1.95Jm346.58 For non stop operation of the part p = 1, result-
ing into O= 1/T. Usage rate of parts for various
= 47 bearings.
organizations could be different. For a petro-
Thus, 47 spares are needed for smooth running leum related industry, typical usage rates for var-
of the system for 2 years, and it will provide 95% ious spare parts of centrifugal pumps are given in
surety that the spares will be available when the Table 3. Such data base could be developed for
need arises. any industry by processing the historical time be-
tween failure data.
The failure intensity can also be calculated by Next step is to develop a link between intensity
a less sophisticated method of using historical of failure 1/ T, usage rate 0, and the demand rate
54

TABLE 3 or
Usage rate of parts of an overhung centrifugal pump of hori- 1 D
zontal type (14)
T NOP
Type of part Usage rate Thus eqns. (6) and (7 ) in terms of calendar
Adapter: pump: 0.0025 I time, t,, and demand rate D can be expressed by
Bearing: ball; 0.05714 using eqns. ( 12 ), ( 13 ) and ( 14) as follows;
Bushing: stuffing box; 0.01019
Casing: 0.00526 N=t,D+0.5(K2-l)+K&?%-‘(p) (6a)
Coupling Half: driver; 0.00227
Coupling half: pump; 0.0054 I
N=t,.D+&ik-‘(p) (W
Cover: bearing; 0.00167
Once the usage data and factors have been ob-
Deflector: 0.01220
Exchanger: heat; assembly; 0.00596
tained, the next step is to develop the appropri-
Frame: pump; 0.00209 ate inventory and order policy consistant with
Gasket: bearing cover; 0.00984 risk of shortage (stockout penalty), inventory
Gasket: casing; 0.05577 investment, criticality of the system availability
Gasket: heat exchanger; 0.01399
(lead time of the part) and location of the pro-
Gasket: seal flange; 0.07833
Gasket: shaft sleeve: 0.03085
cess with respect to supply sources.
Impeller: 0.00804
Lockwasher: impeller; 0.01194 3. Spare part classification
Nut: impeller; 0.00755
Ring: casing: wear; 0.00892
Spare parts need to be evaluated in terms of
Ring: impeller; wear; 0.00776
Ring: lantern; 0.00892
cost and criticality. The cost relates to purchase
Ring: oil: 0.01171 cost and is classified as low, moderate or high.
Ring: retaining: bearing: 0.02302 Criticality is based on the cost of not completing
Shaft: pump: 0.01252 the process or assigned equipment function, i.e.,
Sleeve: shaft: 0.01687
“The Mission”. The cost of not having a part
which shuts down the system or process is one
D of the part. Intensity of failure is a term famil- extreme. The other extreme is that alternatives
iar to reliability engineers, whereas the demand are readily available so that a part loss will not
rate is the term which is much more familiar to stop the continuation of the process. For exam-
the inventory management and material pro- ple, redundant systems are in place.
curement persons. Whereas usage rate U is linked Criticality can also be classified as low, mod-
with demand rate D as follows: erate or high. Highly Critical parts are those
which are absolutely essential for the mission
D=N,(t,)U=N,U (11) success. Moderately Critical parts are such that if
where N,( to) = N,= total number of parts which they are out of stock at the time of demand it will
are in operation in the plant at some reference have a slight to moderate affect on the mission
time to> 0. If we monitor these failures and re- success, whereas parts of low criticality are not
placements for a time period tk (measured from absolutely essential for mission success. If they
origin 0 ), then t, = t L - to, represent a time inter- are not available on demand, alternate parts can
val in terms of calendar year. Combining eqns. be substituted, or in plant manufacturing of such
(9)) ( 10) and ( 11) for N,, parts operational in parts is possible, or they are instantly available
the system; in the market.

Ct=p&=pN,t<. (12) 4. Ordering strategies based on part


and classification

- P
T=B=m=BP NOP Considering cost and criticality of parts jointly,
(13)
one can represent the loss to the organization due
55

TABLE 4 terms of service level (risk) and safety stock need


to be considered. The relevant model for order-
Loss matrix to determine ordering policies
ing strategy 2 is a standard inventory model used
cost Criticality to calculate safety stock associated with a low
probability of stock out. The formula to calculate
Low Moderate High the optimum reorder quantity Q* for this model
Low I,,*** ILM*** 1I_”*** is [6]
Moderate IML** IMM** IM”*
High IHt.** IHM* IHH* 2Dc
I,.,=loss due to the event, that a part of cost i (i=L,M,H ).
and criticality k (k = L,M,H) is out of stock when demanded.
The ordering strategies can be decided on the basis of above
Q*= c J h

loss matrix, by grouping various closely related outcomes to-


where:
gether (indicated by symbols, *, **, and *** in the loss matrix).
C, = Cost of placing an order, $/order
to their absence on demand, in the form of a loss Ch = Annual holding and storage cost per unit of
or regret matrix as given in Table 4. average inventory, $/ (unit, yr )
Ordering Strategy 1: For parts grouped as class
This formula is consistent with the probablis-
1 (symbol*); Set [L, h, bl. tic inventory model illustrated in Figs. 2 and 3.
Maintain a given quantity of these items in the The model’s variables are mentioned in Figs. 2
inventory for compensation of repair times and and 3, illustrate the concept of safety stock, and
procurement lead times, etc, and order new items associated risk of stock out (dark shaded area of
on a one for one basis when failure occurs, and a
spare is withdrawn from the inventory. This
strategy will minimize the risks involved due to
tying up too much capital and the resultant high
I n
inventory maintenance cost. Initial stock level for
such parts can be calculated using eqns. (6) or
(7) depending upon the coefficient of variation :\;;,+/!Y-f
\
K and incorporating a high value of p. For ex-
ample a value of p = 0.999, implies that there is a
very low probability of stockout in time interval

O-t,, where t, is much less than the mission Rraton&lc\’
Maximum \I Safety
time, but slightly greater than Ornand'\ I Stock
[ F+ 3~+ (Manufacturer’s lead time + Shipment \I I
I 1
i I I 1 I _
time) ] i.e.
t, > [ T+ 30+ (Manufacturer’s lead time The

Fig. 2. Illustration of inventory model and related variables.


+ Shipment time) ]
For additional factor of safety the number of parts
calculated using eqn. (6 ) or ( 7 ) corresponding
to t = t,, can be increased by one in the initial ac-
quisition. Later on the orders are placed on one d
by one basis.
Ordering Strategy 2: For parts grouped as Class
2 (symbol**); Set ]&, lMM,Ll
In addition to the purchase cost side of the de-
cision process, the other critical cost factors in Fig. 3. Effect of normal demand during lead time.
56

the normal probability curve in Fig. 3). Average Example 2


demand during lead time (LT) is calculated from
eqns. ( 1) and (4) and standard deviation of de- In an existing inventory system the following
mand during lead time is calculated from eqn. information is available for a part of moderate
( 3 ) or ( 5 ) by replacing t = LT in these equations. cost and low criticality:
The effect of high stockout costs will increase
the order quantity, which decreases the fre- Existing inventory policy
quency of exposure to risk of stockout. If the cost Item: filter assembly. Code: XY-WX-YZA.
of stock out C, can be assigned to the number of On hand inventory: 28, Safety stock = 0,
order cycles (exposures to a stockout) then the Unit cost: c,,= 128 $ (moderate )
formula for the optimum order quantity can be Holding cost: 15% of 128 $ = 20 $
modified as Average Demand Rate: D= 13 parts/year
Planning Horizon: = 1 year
Coefficient of Variation of Life: K= 1

J
2o(c,+c,)
Q*=
Ctl Proposed inventory policy
The number of spare parts needed in one cal-
Ordering strategy 3: For parts grouped as Class
endar year with a risk of being out of stock equal
3 (symbol***); Set [ lLL, lLM, ILH ]
to 0.05 (i.e., p=O.95), can be calculated using
These parts are of low cost and varying degree eqn. (6a):
of criticality. There are two possible strategies for
N=lx13+@0-‘(1.65)
this type of parts:
Strategy 3(a): If the following conditions are = 13+6= 19 parts.
simultaneously satisfied: Refering to the loss matrix, the inventory deci-
(i) spare parts can be stored for an entire mis- sion correspond to a value lLM, which means
sion duration without any damage during strategy 3 (b) need to be adopted, i.e., using eco-
the storage; nomic order quantity formula and adding some
(ii ) storage cost for these parts is very small; safety stock. For a one year inventory cycle
(iii) there is a danger of non availability of
these parts during the mission time; and

J
2x13~128
(iv) parts are not instantly accessible in the lo- Q*= = 12.89= 13 parts
20
cal market, they have to be ordered suffi-
ciently ahead of time; SS=fi( 1.65)=6 parts.
then the strategy for ordering is to order them in
one lot of N items to cover the entire mission pe- Thus the proposed policy is to initiate an order
riod. For a specific risk of stockout say of 13+6= 19 parts, and if the lead time is as-
1 -p=O.O5, or p=O.95, and t representing the sumed to be 8 months, then place an order of 19
mission time, the order quantity Q=N can be items when the inventory level reduces to 15 parts
calculated using eqn. ( 6 ) or ( 7 ) . (Q,=13~8/12=8.63=9parts+asafetystockof
Strategy 3 (b) : If one or more of the above con- 6 parts). The existing policy is to have zero safety
ditions is not satisfied then make several smaller Prostock and maintaining 28 items per year. The
orders during the mission period, which will be proposed policy demonstrates that the parts are
identical to the strategy 2, discussed earlier. 47% overstocked. If the part has KC 1 and/or the
The ordering strategies discussed above cover planning horizon is more than one year, then N,
both the safety stock and initial stocking level. calculated by eqn. (7a) or (6a), will differ from
However, the costs and risks need to be balanced 19 parts. The Q* will remain unchan ed, but the
in order to arrive at an inventory stocking level safety stock will adjust to SS=K ,B 13 ( 1.65) =
for specific spare parts. The following section K( 6). For example for K=0.3, the safety stock
summarize the overall decision process in setting will reduce to 0.3 x 6=2 parts. This demon-
industrial policies. strates the influence of K on the safety stock.
57

Discussion and summary sessment of industrial plant locations and pro-


cess in order to make proper decisions. Some-
times the cost of a shutdown outweights any other
Having assessed the cost of ordering and the cost and a redundancy of equipment policy may
cost of stockout based on safety stock analysis, a become superior to spare part stocking policy.
decision process can be developed to balance the The concepts expressed in this paper are to
results and determine a spare part policy. This provide a formal method of determining spare
process can take the form of the following deci- part stocking levels which in industry today tend
sion rules: to be overstocked in most plants, are moving
(a) If purchase cost is low and critical cost is slowly (low turnover rate), and require a high
low to high and the part is readily available cost of stockage. Less formal methods of spare
then maintain stock at local supplier or use part inventory policy have not worked. What is
just in time techniques. needed is a well planned, integrated methodol-
(b) If purchase cost is low and critical cost is ogy beginning with equipment population, fail-
low to high and the part is not readily avail- ure costs (usage) and ordering/inventory cost
able then order large quantities of stock for conditions. Applying them together and properly
life of mission. will result in minimal inventory with higher con-
(c) If purchase cost is medium and criticality is fidence levels of reduced risk.
low to medium or purchase cost is high but
criticality is low, then order in batches at References
optimal order quantity and maintain a rea-
sonable safety stock. Czichos, H., 1978. Tribology - A Systems Approach to
(d) If purchase cost is high and criticality is me- the Science of Technology of Friction, Lubrication and
Wear. Elsevier, Amsterdam, p. 240.
dium to high, maintain an absolute mini- Gnedenko, B.V., Belyayev, Yu.K. and Solovyv, A.D., 1969.
mum+ 1 additional spare, and order fre- Mathematical Methods of Reliability. Academic Press,
quently on one by one basis. New York.
Material availability and location of the pro- Lipson, C. and Sheth. N.J., 1973. Statistical Design and
Analysis of Engineering Experiments. McGraw-Hill, New
cess from supply sources will play a significant
York.
role in shaping the inventory policy. Fluctua- Bloch, H.P. and Geitner, F.K., 1983. Practical Machinery
tions in failure rates and population of equip- Management for Process Plants, Vol. 2. Gulf Publishing
ment size will need to be mentioned to assure that Company, Houston.
proper inventory levels are established. Sheikh, A.K., 1990. A statistical method of calculating the
spare part requirement. Conf. Proc.Symposium on Main-
Historical data of part usage will provide in- tenance, Planning and Operations, 17- I9 March 1990 at
formation to prevent the stocking of parts which King Saud University, Riyadh, Saudi Arabia pp. 204-222.
never fail or can be repaired to their original state. Monks, J.G., 1985. Operations Management. McGraw-
The balance of the various costs require an as- Hill Book Company.

You might also like