A Kohonen self-organizing map approach to addressing a multiple objective,mixed-model JIT sequencing problem

A technique is presented which addresses a JIT production-scheduling proble m where two objectives are present - minimization of setups between differi ng products and optimization of schedule flexibility. These two objectives are inversely related to each other, and, as a result, simultaneously obtai ning desirable results for both is problematic. An efficient frontier appro ach is employed to address this situation, where the most desirable sequenc es in terms of both objectives are found. Finding the efficient frontier re quires addressing the combinatorial complexity of sequencing problems. The artificial neural network approach of a Kohonen self-organizing map (SOM) i s used to find sequences which are desirable in terms of both the number of setups and flexibility. The Kohonen SOM was used to find sequences for sev eral problems from the literature. Experimental results suggest that the SO M approach provides near-optimal solutions in terms of the two objectives, in addition to comparing formidably with other search heuristics. Results a lso show, however, that the SOM approach performs poorly with regard to CPU time. (C) 2001 Elsevier Science B.V. All rights reserved.