Cellular manufacturing is well known as an effective way to improve worksho
p performances. There are various methods to design cells. Most of them do
not take into account constraints specific to the workshop. We are interest
ed in the design of manufacturing cells that can take into account specific
constraints (for example, certain machines may have to stay together in th
e same cell because they will share a common resource or certain machines m
ay have to be separated because they will produce interferences). The propo
sed method uses evolutionary algorithms. The initial solutions are created
with an algorithm based on a random tree search and then the solutions evol
ve thanks to operators designed so as to satisfy, at any stage, the constra
ints. The suggested method is illustrated through a problem with a known op
timum in order to verify that the evolutionary algorithm can find the optim
um solution. (C) 2000 Elsevier Science B.V. All rights reserved.