Clustering based algorithms have been used for machine cells grouping for years and have been considered as a feasible approach because the algorithms are flexible and easy to implement on computers. However, two major deficiencies have been identified for clustering algorithms in this research; they are inconsistency and possible mis-clustering. Inconsistency is usually caused by arbitrary determination of initial cluster centres and mis-clustering, from the manufacturing point of view, is due to pattern similarity as the unique criterion for the cluster-seeking process.

A new heuristic clustering algorithm has been developed and applied to formation of machine cells. The algorithm not only overcomes common drawbacks of clustering methods, but also minimizes bottleneck machines required. The algorithm consists of three parts: 1) determination of initial clustering centres, 2) cluster-seeking process, and 3) minimization of number of bottleneck machines. The initial clustering centres are determined by finding a set of components which require the most different machine operations, then the cluster-seeking process is carried out based on manufacturing similarity. The final step is to minimize the number of bottleneck machines required for forming independent cells. Another advantage of the algorithm is that it can generate alternate cells configurations by simply changing input parameters. This provides an opportunity to resolve conflict between formed cells and some physical constraints such as plant space, machine location, cranes available and the like. Examples are provided to illustrate the approach.

This content is only available via PDF.
You do not currently have access to this content.