As part of the cellular manufacturing design process, machines must be grou
ped in cells and the corresponding part families must be assigned. Limits o
n both the number of machines per cell and the number of parts per family c
an be considered. A weighted sum of intracell voids and intercellular moves
is used to evaluate the quality of the solutions. We present a tabu search
algorithm that systematically explores feasible machine cells configuratio
ns determining the corresponding part families using a linear network flow
model. The performance of this tabu search is benchmarked against two simul
ated annealing approaches, another tabu search approach and three heuristic
s: (ZODIAC, GRAFICS and MST).