In this paper, a methodology for the design of a flexible assembly lin
e system (e.g. a flow shop consisting of a set of tandem flexible work
stations, a set of automated inspection stations, a loading station, a
nd an unloading station, linked by a material handling system) is pres
ented, while controlling the botteneck problem. This paper is a genera
lization of earlier work1, where a design strategy far a simplified ve
rsion of a similar manufacturing system was proposed. In this paper, t
he loading process is stationary with identically distributed interarr
ival times, and defective work-pieces are reprocessed at the correspon
ding workstations. Here, the performance of the assembly line is model
led by a non-Markovian queueing network with finite capacity queueing
nodes. Based on this queueing network, a stochastic optimization model
is then presented to select the minumum required local storage sizes
of the workstations and the transporter stations such that the resulti
ng probabilities of finding each workstation and each transporter stat
ion blocked (e.g. full) are sufficiently close to zero. To solve the p
roposed optimization model, a heuristic algorithm is then developed. F
inally, an example is presented, and the relative accuracy of the prop
osed algorithm is investigated by a simulation study.