Wen et al. [104] proposed a dynamic routeing method   using a fuzzy part-family formation approach, combined with a cer- tainty factor procedure, to suggest the  best  route  in  a multi- cell FMS. In order to take into account the quantitative estimate of the relationship between machines and parts, a fuzzy clus- tering algorithm was suggested. The clustering algorithm was based on the distance from the clustering centre, a smaller distance associated with a higher degree of membership. The fuzzy clustering approach not only revealed the specific part family that a part belongs to, but also the degree of membership of a part associated with each part family. In  summary, the fuzzy clustering algorithm provided extra information that was not available in conventional algorithms. The following rules were some  examples of the routeing  decision   process:

If “part type 1 enters cell-1 for processing” then “the degree of  difficulty  is 0.4”.

If “part type 1 enters cell-2 for processing” then “the degree of  difficulty  is 0.96”.

A simulation model was constructed to compare the perform- ance of the proposed dynamic routeing method with the per- formance of the fixed routeing method. The model was developed using SLAM II, which consisted of two machine cells, 22 part types, and two  AGVs  (each  machine  cell had one AGV). The dispatching rule used in the model was FCFS. Three performance measures were employed consisting of mean flow-time, mean tardiness, and absolute lateness. The perform- ance measures were obtained under  three  different  levels  of the  system utilisation.

Chan et al. [105] developed a fuzzy approach for operation and routeing selection in an FMS via simulation. The FMS consisted of six workstations, a finite input and output buffer at each station, a load/unload station, and three AGVs. The authors used a fuzzy approach to  study  operation  selection first. It was compared with five operation selection rules, which were RAN, SNQ, LULIB, CYC, and WINQ. Performance measures employed were net profit, makespan, average lead time, average tardiness, average lateness, average machine utilisation, average WIP at the input buffer, and average delay at the local buffer. Results showed that the proposed method performed better than the other rules on the performance measures other than makespan, and average WIP at  input buffer. The authors then applied the fuzzy approach to routeing selection. It was compared with three rules, which were SNQ, WINQ, and SPT. Performance measures were average and maximum flow-time, average and maximum tardiness, average and maximum lateness, number of tardy jobs, number of completed jobs, and net profit. Results showed that the fuzzy approach outperformed the other selection rules on all the performance measures. The authors concluded that the proposed method showed a good improvement in some performance mea- sures.

In recent years, genetic algorithms (GAs) have received significant attention because of their special evolutionary mech-

anism. GAs have also been used to solve FMS scheduling problems. Fang and Xi [106] studied a hybrid of GAs and dispatching rules for solving the job shop scheduling problem with sequence-dependent set-up time and due-date constraints. They concluded that the proposed strategy is more suitable   for a dynamic job shop environment than the static scheduling strategy.

Jawahar et al. [107] proposed a GA to derive an optimal combination of priority dispatching rules. The performance was compared for makespan criteria and computational time of an FMS. The proposed algorithm was compared with four dis- patching rules, which were SPT, LPT MINSLK, and EDT. It was shown that computational time obtained from the GA approach was less than that from the other methods but, for most of the time, it did not provide the optimal  solution. Jawahar et al. [108] then carried out work on two knowledge- based scheduling schemes (work cell attribute oriented dynamic schedulers “WCAODSs”) to minimise the makespan perform- ance criterion. Results were compared with GA-based method- ology and the results indicated that there was no significant variation in the solution output. The same FMS was employed both times,  consisting of 2 to 10  work cells, which  was  either a machining centre or an assembly machine or an inspection station. An AGV was used to transfer materials, pallet, and fixtures  between  work cells.