Rd. Meller et Ya. Bozer, ALTERNATIVE APPROACHES TO SOLVE THE MULTI-FLOOR FACILITY LAYOUT PROBLEM, Journal of manufacturing systems, 16(3), 1997, pp. 192-203
Citations number
16
Categorie Soggetti
Engineering, Manufacturing","Operatione Research & Management Science","Engineering, Industrial
Alternative approaches to solve the multi-floor facility layout proble
m are compared in this paper. In particular, because the multi-floor f
acility layout problem is NP-complete, the effectiveness of heuristica
lly solving it via a single-stage approach (where each department, unl
ess it is fixed or constrained by the user, is allowed to occupy any f
loor during the execution of the heuristic) versus a two-stage approac
h (where in the first stage each department is permanently assigned to
one of the floors and in the second stage the layout is determined fo
r each floor) is evaluated. Generally speaking, given no limits on run
times, one would expect the single-stage approach to outperform the a
bove two-stage approach because the former allows departments to chang
e floors during the execution of the algorithm. However, if the run ti
mes are required to be reasonable and comparable, it is no longer clea
r that the single-stage approach will lead to superior solutions becau
se it has to explore a larger solution space. The performance of the t
wo-stage approach, on the other hand, depends on the quality of the fl
oor assignments obtained in stage one. Instrumental to the quality of
these assignments is a newly developed simplified formulation with a l
inear objective to optimize the inter-floor handling cost in stage one
. The resulting two-stage approach is then evaluated with respect to a
single-stage approach using various test problems. The evaluation inc
ludes a variation of the above two-stage approach where the floor assi
gnments obtained in stage one are not permanent; that is, departments
are allowed to change floors in the second stage. The evaluation indic
ates that the proposed two-stage approach compares quite well to both
of the other approaches. The results obtained by the two-stage approac
h are shown from a data set that is based on a real-world manufacturin
g facility.