Work In Next Queue CONWIP

Abstract

The present paper is aimed at showing, through a simulative approach, that the adoption of a suitable dispatching rule allows to improve the single-loop CONstant Work In Process (CONWIP) control mechanism within Make To Order (MTO) production systems, balancing the workload among the workstations, reaching a performance level that outperforms standard CONWIP and leans towards that of the corresponding m-CONWIP system. The benefits that may derive from the adoption of a single-loop CONWIP for the design/management of production systems are obvious, being true that (i) m-CONWIP systems are complex to design and optimize, (ii) the single-loop CONWIP systems can be designed with simpler approaches and that (iii) single-loop CONWIP systems remain easier to manage than the m-CONWIP systems. Thus, the well-known Work In Next Queue (WINQ) rule has been adjusted and used within a single-loop CONWIP model, to guarantee that items are favored within those routings for which higher capacity is available in the succeeding work centers. Its performance has been then compared respectively to that of the standard, FIFO-based, CONWIP system and that of an extremely efficient m-CONWIP system. It is known that the workload balancing capability of pull systems not only depends on the configuration of the system itself. It is also subjected to the variability in the order arrival pattern and of the processing times. These parameters have been therefore opportunely considered. Also, the number of available cards represents, along with the loading rule, the most important control parameter. It can be easily determined both in the standard and the WINQ-based CONWIP, whereas it represents a significant issue within the m-CONWIP systems. Thus, to optimize the number of cards within the m-CONWIP model a Genetic Algorithm (GA) has been opportunely configured.