An effective teaching learning based optimization for flexible job shop scheduling

Abstract

A flexible job shop scheduling problem (FJSSP), a prolongation of former job shop scheduling problem (JSSP), is well known as a NP-hard (Non-deterministic polynomial time) problem. The FJSSP contains the difficulties of classical JSSP with an additional complexity of allotment of operations to the machines. A FJSSP has two sub problems. They are (i) routing sub problem i.e. allotting of each operation to a machine from a given set of machines capable of doing it and (ii) sequencing sub problem i.e. arranging of allotted operations on each machine. It is not always possible to get an optimal solution to such problems in a reasonable computation time owing to large computational complexity involved. Therefore, a variety of meta-heuristic approaches have been used to obtain near optimal solutions in a reasonable computation time. Teaching learning based optimization (TLBO), which needs minimum number of parameters to be tuned in comparison to other meta-heuristic approaches, can be an efficient algorithm. Like many meta-heuristics, TLBO may get trapped at the local optimum and present difficulty in reaching the best solution. In this paper, this limitation is addressed efficiently by embedding TLBO with a new effective local search technique. Real number encoding system is used and the FJSSP is solved based on integrated approach. Tests are carried out on famous bench mark problems with minimization of makespan as the objective. Results obtained from modified TLBO have been compared with other algorithms from literature.