Advances in Pure Mathematics, 2012, 2, 128-132
http://dx.doi.org/10.4236/apm.2012.22019 Published Online March 2012 (http://www.SciRP.org/journal/apm)
Reliability Analysis of Systems Based on the UFLP under
Facility Failure and Conditional Supply Cases
Min Wang, Zongtian Wei, Yun He
Department of Mathematics, Xi’an University of Architecture and Technology, Xi’an, China
Email: ztwei@xauat.edu.cn
Received November 24, 2011; revised January 10, 2012; accepted January 20, 2012
ABSTRACT
The reliability of facility location problems has been received wide attention for several decades. Researchers formulate
varied models to optimize the reliability of location decisions. But the most of such studies are not practical since the
models are too ideal. In this paper, based on the classical uncapacitated fixed-charge location problem (UFLP) and
some supply constraints from the reality, we distinguish deterministic facility failure and stochastic facility failure cases
to formulate models to measure the reliability of a system. The computational results and reliability envelopes for a
specific example are also given.
Keywords: Reliability Analysis; Facility Failure; Supply Constraint; Uncap acitated Fixed Charge Location Problem
1. Introduction
The uncapacitated fixed-ch arge location problem (UFLP)
[1] is a classical facility location problem that chooses
facility locations and assignments of customers to facili-
ties to minimize th e sum of fixed and transpor tation costs.
Once a set of facilities has been constructed, however,
one or more of them may from time to time become un-
available, we call it the facility failure. For ex ample, due
to inclement weather, earthquakes, sabotage, or changes
in ownership. The facility “failures” may result in exces-
sive transportation costs as customers previously served
by these facilities must now be served by more distant
ones. Snyder and Daskin [2] define the “reliability” of a
system as the ability of a system to perform well even
when parts of the system have failed. They formulate
models based on the UFLP and other classical discrete
location problems to optimize the reliability of a system.
However, they ignore an actual situation that suppliers
would not like to serve the customers too far from them
when the supplier who serves these customers previously
has failed.
Throughout this pa per, we use the concept “reliab ility”
defined in [2]. In this paper, ad opting th e facility locatio n
analysis framework, we consider facility systems’ reli-
ability analysis based on the UFLP and some supply cons-
traints when subject to facility failu res.
The remainder of this paper is structured as follows.
We formulate two reliability analysis models based on
the UFLP in§2. The computational results and the reli-
ability envelops of an example are presented in§3. In§4,
we give a summary of this paper.
2. Models
The facility location problem is a classical optimization
problem to determine the number and locations of a set
of facilities and assign customers to these in such a way
that the total cost is minimized. Two types of costs are
considered in the problem. A setup cost (facility cost) oc-
curs while a facility is opened, and a connection cost
occurs while a customer is assigned to the opened facility.
A fruitful research results have obtained in this field and
the topic remain hot still (see [3-6] and the references
therein).
2.1. The Uncapacitated Fixed-Charge Location
Problem
If an arbitrary number of customers can be connected to
a facility, the problem is called UFLP. The follows is th e
model of the classical facility location problem UFLP.
Sets:
I
: set of customers, indexed by ; i
J
: set of potential facility locations, indexed by . j
i
hiI
Parameters:
: demand at customer ;
f
: fixed cost of open a facility at ;
j
jJ
ij
dij
i
c: the distance from customer to facility ;
: transportation cost per un it per distance;
: weight of the fixed cost in the objective function.
Decision variables:
C
opyright © 2012 SciRes. APM
M. WANG ET AL. 129
1,ifa facilityisopened
0, ot
j
X
atlocation
herwise
j
dtofacility
herwise
ij
1, if customerisassign
0, ot
ij
Y
The model of the UFLP is as follows

min 1
j
j
jJi iiijij
IjJ
Z
fX



chdY



(2)
(3)
(4 )
(5)
*
(1)
1
,
.. 0, 1
0,1 ,
ij
jJ
ij j
j
ij
YiI
YX iIjJ
st XjJ
YiIjJ




The objective function (1) minimizes the weighted
sum of the fixed cost for opening the facilities and the
demand weighted total distance. Constraints (2) stipulate
that each customer is assigned to exactly one facility,
while constraints (3) limit assignments to open sites. Fi-
nally, constraints (4) and (5) are integrality co nstraints.
2.2. The Reliability Analysis Model under
Deterministic Facility Failure Case
In the UFLP, when o ne or mo re facilities have failed, the
overall supplement of the facility system will decrease
dramatically but the total demand does not change. If in
this situation all demands must be met, then every node
must has no any supply capacity limit. However, the ca-
pacity of facilities are designed a priori, when facility
failures happen, how can the remained facilities guaran-
tee the increased demands? Even the remaining facilities
can satisfy the whole demands, the transportation cost
will increase dramatically because the customers served
by the failed facilities now must be served by the re-
maining facilities that are far away from them. In most
cases, conditional service is a desirable policy, e.g., once
the weighted distance between a supplier and a customer
exceed some value, the demand can be given up. Based
on this idea, we formulate a new reliable model under
deterministic facility failures as the follows.
Let
j
and ij
Y be an optimal solution of the UFLP
and S be the corresponding system. Denote
*
*
C:

1
j
jX

and :1
jij
NiY

,1,2,,.j C
Let
F
C be the potential failure facility set, where
the failure is defined as a facility losses its designed
function completely. Let r be the set of scenarios cor-
responding to the failure of r facilities from S, i.e., every
r
S
s
Sr explicitly specifies the failed facilities in S.
Denote the failed facility set corresponding to
s
as
s
F
,
then the set of customers which need to reassign suppli-
ers is
s
j
s
jF
NNF
r
.
Parameters:
: number of failed facilities;
j
Vj
: supply constraint of facility ;
i
: per unit penalty due to give up the supply to cus-
tomer ;
i
i
c i: transport cost to customer per unit per distance.
Max ,
j
jiiji
iN
VhdjC
Where

,cdi N
and
iiiji

, i
and i
are the weight ratio,
and ij denotes the shortest distance between customer
i and distribution center such that Y in the op-
timal solution of the UFLP.
dj1
ij
1,
Y
We define the assignment variables as ijs
if cus-
tomer i is served by facility jin scenario
s
, ,

0other-
wise. The model is:


\\
Min min1
1
r
sjs
rj
sS jC
i iij ijsi iijsi iij
iNjCFiN jCF
Zf
chd YhYchd

 






 
\
..1, ,
s
ijs r
jCF
(6)
s
tY iNsS

(7)
,,\,
iij ijsjsr
hd YViNjCFsS (8)
 
0,1 ,,\,
ijss r
YiNjCFsS (9)
 
r
r
The objective function (6) selects failed facilities
from F in order to minimize the resulting total cost. Con-
straints (7) require that each customer be served by at
most one server in any scenario. Constraints (8) represent
the supply conditions. Constraints (9) require the as-
signment variables to be binary.
Changing “Min” to “Max” in the objective function,
then we can obtain the worst case model that is the model
to measure the maximal system cost under the facility
failure level .
2.3. The Reliability Analysis Model under
Stochastic Facility Failure Case
The reliability model formulated above is based upon a
deterministic analysis. We now consider the case where
the facility failure is not a certainty. Usually, the chances
of losing a facility are based upon some probability. We
wish to derive the maximal or minimal expected effi-
ciencies associated with an existing system. To do this
we need to identify both the worst case and the best case
expected outcome s.
F
Let E be the target facility (the potential failure
facility) set of an attack. Assume that an attacker can hit
Copyright © 2012 SciRes. APM
M. WANG ET AL.
130
F
at most once and that the facilities in each facility in
F
will be hit simultaneously. Let be the scenario
set when r
S

0rrF facilities in
F
have been at-
tacked. Each r
F
sSr specifies which facilities in
F
have been attacked. We also use
s
to denote the
facility set that have been attacked. Then any s
F
s
can be used to represent a failed facility set in scenario
s
. Let
j
p be the failure probability of facility jF
after one attack and the failures are independent each
other for any two facilities. It is easy to see that scenario
s
F
occurs with probability

\
1
ss
s
F
jj
p
Y
jF j

sF
1,
Pp
We define the decision variables as s
ijF if cus-
tomer is served by facility in scenario
i j
s
F
; 0,
otherwise. The parameters not define here are as the
same in that of the last subsection. The model is as the
follows.


\
\
min
ss
ss
js
rj
jC
FijF ijF
iNjCF
iiij
jCF
Zf
h Y
chd





r
1
s
i i
dYh

,
Min
r
sS
Fs
iN
1
i i
pc
1,
ij

(10)
\
.. s
s
ijF
jCF
s
tY

iNs
\
s
C F
\ ,
s
Nj
r
S
,
r
s S
r
CFsS
r
r
,
(11)
,,
s
j ijFj
Y Vj

0,1 ,,
i i
hd
ijF
i
Yi
N (12)
s (13)
The objective function (10) selects failed facilities
from F in order to minimize the resulting total cost ex-
pectation. Constraints (11) require that each customer be
served by at most one server in any scenario. Constraints
(12) represent the supply conditions. Constraints (13)
require the assignment variables to be binary.
Changing “Min” to “Max” in the objective function,
then we obtain the worst case model, that is, the model to
measure the maximal system cost expectation under the
facility failure level .
3. A Computing Example and Reliability
Envelopes
The models described above can be applied to a given
facility system over a range of facility loss level . One
can easily enumerate each of the possible ways of losing
one facility as well as calculate the impact of each possi-
ble loss in terms of changes in cost. The results of this
series of calculations will define a range of losses from
the best case (i.e. the least increase in cost) to the worst
case (i.e. the greatest increase in cost). We then have a
region defined by an upper curve and a lower curve,
where the upper and the lower curve represent the solu-
tions of the least or the greatest impact associated with a
given facility loss level, respectively. The region de-
picted between these two curves can be defined as the
operational envelope or reliability envelope. For a given
edge loss level, this envelope specifies the range of pos-
sible system performance from the best-case to the
worst-case. Actual performance will fall within this
range.
J
I
Assume
we use the data set (see [7]) to opti-
mally solve an UFLP with 0.7
1.2
in order to establish
a facility system. Figure 1 shows the optimal solution,
where the 8 distribution centers are city 4, city 7, city 11,
city 20, city 24, city 26, city 28 and city 45, and the
edges marked by red color represent the delivery routes
from each distribution center to its customers.
Given this operating system of 8 facilities and a poten-
tial facility failure set F which is consisted of 7 facilities:
4, 7, 11, 20, 24, 28, 45, and i1.5, i
, we
solve the deterministic facility failure model. The solu-
tions are given in Table 1.
Since the 8 distribution centers are also customers and
city 26 only serves itself, we let the supply constraint of
city 26 be
max,and26.VVjCj
i
j The penalty
value
of the customers in are established to be C
max ,\ .iIC


i
By using the same data, we then solve the stochastic
reliability model with facility failure probability p = 0.5.
The solutions are shown in Table 2. In this paper, we
define the reliability of a system as the ratio of the sys-
tem’s total operational cost when no facility failures and
the total cost after some facilities have failed. Figures 2
and 3 are the corresponding reliability envelopes.
Figures 4 and 5 give the operational conditions of the
Figure 1. Optimal solution of the UFLP with α = 0.7.
Copyright © 2012 SciRes. APM
M. WANG ET AL.
APM
131
Table 1. Solution of the deterministic reliability model with α = 0.7.
Level Best-Case Worst-Case
r Objec. Value Failed Facilities Efficiency Objec. Value Failed Facilities Efficiency
0 31890.22 - 1 31890.22 - 1
1 33139.68 24 0.9623 41764.52 7 0.7636
2 34617.96 24 28 0.9212 49382.51 7 11 0.6458
3 36893.28 24 28 45 0.8644 60474.09 4 7 11 0.5273
4 40725.83 4 11 24 45 0.7830 70701.22 4 7 11 45 0.4511
5 44537.72 4 11 24 28 45 0.7160 75905.44 4 7 11 28 45 0.4201
6 48673.00 4 11 20 24 28 45 0.6552 79165.84 4 7 11 20 28 45 0.4028
Table 2. Solution of the stochastic reliability model with facility failure probability 0.5.
Level Best-Case Worst-Case
r Objec. Value Failed Facilities Efficiency Objec. Value Failed Facilities Efficiency
0 31890.22 - 1 31890.22 - 1
1 32514.95 24 0.9808 36827.37 7 0.8659
2 33254.09 24 28 0.9590 40636.36 7 11 0.7848
3 34391.75 24 28 45 0.9273 44618.43 4 7 11 0.7147
4 36309.65 20 24 28 45 0.8783 47106.28 4 7 11 45 0.6770
5 38941.59 4 20 24 28 45 0.8189 48536.07 4 7 11 20 45 0.6570
6 42755.11 4 7 20 24 28 45 0.7459 49951.31 4 7 11 20 28 45 0.6384
Figure 2. The reliability envelope associated with solutions
presented in Table 1. Figure 3. The reliability envelope associated with solutions
presented in Table 2.
system when facility failure level is 3, since the probabil-
ity of this case is the largest one. system when subject to facility failures. We distinguish
deterministic and stochastic cases to formulate and com-
pute a specific example. Reliability envelopes in these
two different cases are also given. The information in the
reliability envelopes can be very useful in looking at
ways to protect a facility system.
4. Summary and Conclusions
In this paper, we propose two types of scenario based
models in order to analyze the reliability of an existing
Copyright © 2012 SciRes.
M. WANG ET AL.
132
Figure 4. The Best-Case with 3 fail e d facilities.
Figure 5. The Worst-Case with 3 failed facilities.
Therefore, the value of our analysis could lead to
higher levels of safety as well as efficient levels of re-
source allocation for security measures (whether that
involves a possible natural disaster or an attacker).
We also notice that, when the facility loss level r = 3
in the deterministic model, the optimal solution in the
worst case are not very ideal, since the transport distance
of facility 24 are too far, so the service time will be too
long. We need to do some improvement in our models,
i.e., the service and other constraints.
5. Acknowledgements
This paper was supported by the SXESF (No. 09JK545)
and the BSF (No. JC0924).
REFERENCES
[1] M. L. Balinski, “Integer Programming: Methods, Uses,
Computation,” Mana gement Science, Vol. 12, No. 3, 1965,
pp. 253-313.
[2] L. V. Snyder and M. S. Daskin, “Reliability Models for
Facility Location: The Expected Failure Cost Case,”
Transportation Science, Vol. 39, No. 3, 2005, pp. 400-
416.
[3] Z.-J. Shen, “Integrated Supply Chain Design Models: A
Survey and Future Research Directions,” Journal of In-
dustrial and Management Operation, Vol. 3, No. 1, 2007,
pp. 1-27.
[4] J. Current, M. Daskin and D. Schilling, “Facility Location:
Applications and Theory,” In: Z. Drezner and H. W. Ha-
macher, Eds., Springer-Verlag, New York, 2001.
[5] M. S. Daskin, “What You Should Know about Location
Modeling,” Naval Research Logistics, Vol. 55, No. 4,
2008, pp. 283-294. doi:10.1002/nav.20284
[6] L. V. Snyder and Z.-J. Shen, “Managing Disruptions to
Supply Chains,” National Academy of Engineering, Vol.
36, No. 4, 2006, pp. 39-45.
[7] Z. T. Wei and H. Y. Xiao, “Reliability Analysis of Facil-
ity Systems Subject to Edge Failures: Based on the Un-
capacitated Fixed-Charge Location Problem,” Open Jour-
nal of Discrete Mathematics, Vol. 1, No. 3, 2011, pp.
153-159. doi:10.4236/ojdm.2011.13019
Copyright © 2012 SciRes. APM