The effective application of JIT cannot be independent from other key components of a lean manufacturing system or it can “end up with the opposite of the desired result”1. Specifically, leveled production (heijunka) is a critical factor. The leveling problem in JIT, a mixed-model scheduling problem, was formalized in 1983 and named MMJIT. Several numbers of solving approaches for MMJIT have been developed during the last decades. Most of them assume constant demand and product mix. Zero setup-times hypothesis has been removed only since 2000, and few approaches still cope with stochasticity. On top of this, these algorithms, although heuristic based, usually spend too much time in finding a good solution. Simplification hypotheses, operations research competences requirements and slow execution prevented these approaches to widespread in industry. Indeed, the heijunka box or the standard FIFO kanban approach with the simple Goal-Chasing-Method heuristic are still the most used tools to manage production in JIT environment. This is acknowledged also by the proponents of alternatives, and GCM is always used as a benchmark for every new MMJIT solution. However, these traditional approaches are not so effective in case of long setups and demand variations, given the fact that they have been conceived in pure JIT environments. In high stochastic scenarios, in order to prevent stock-outs, kanban number is raised along with the inventory levels. There are several cases of companies, operating in unstable contexts and where setup times cannot be reduced over a certain extent, that are interested in applying JIT techniques to reduce inventory carrying costs and manage the production flow in an effective and simple way. The development of kanban board / heijunka-box variations, in order to cope with the specific requirements of these companies, seems to offer better potentialities if compared to the development of difficult operations research algorithmic approaches. In order to solve industrial problems, researchers may concentrate in finding new policies that could really be helpful for production systems wishing to benefit from a JIT implementation but lacking in some lean production requirements, rather than studying new algorithm for MMJIT problem.
For instance, kanban board / heijunka-box variations can effectively focus on job preemption opportunities in order to reduce setups abundance, or on new rules to manage priorities in case of breakdowns or variable quality rates. The parameters fine-tuning can be performed through simulation. In this sense, given the limitations of most commercial software, the development of a simulation conceptual model – along with its requisites – of a model representation (objects and structures) and some communication rules between the subsystems (communication protocols) are the main issues that need to be addressed from academics and developers.
- 1086 reads
