Assumed that time wastes are a clear example of MUDA in Lean Production1, complex mathematical approaches which require several minutes to compute one optimal sequence for MMJIT 2 should be discarded, given that the time spent calculating new scheduling solutions does not add any value to products. On the other side, it is notable that MRP computation requires a lot of time, especially when it is run for a low-capacity process (in which CRP-MRP or capacitated MRPs are required). However, being MRP a look-ahead system which considers the demand forecasts, its planning is updated only at the end of a predefined “refresh period”, not as frequently as it may be required in a non-leveled JIT context. MRP was conceived with the idea that, merging the Bill-Of-Materials information with inventory levels and requirements, the production manager could define a short-term work plan. In most cases, MRP is updated no more than every week; thus, an MRP run may also take one day to be computed and evaluated, without any consequences for the production plan. On the contrary, the situation in JIT environment evolves every time a product is required from downstream. While MRP assumes the Master Production Schedule forecasts as an input, in JIT nobody may know what is behind the curtain, minute by minute.
Indeed, while a perfect JIT system does not need any planning update – simply because in a steady environment (e.g. heijunka) the optimal sequence should always be almost the same, at least in the medium term – real-world short-term variations can deeply affect the optimality of a fixed schedule production. For instance, a one-day strike of transport operators in a certain geographical area can entirely stop the production of a subset of models, and the lack of a raw material for one hour can turn the best scheduling solution into the worst. On top of this, while MRP relies on its “frozen period”, JIT is exposed to variability because is supposed to effectively react to small changes in the production sequence. However, some authors noticed that the JIT sequences3, 4, 5 ] are not so resistant to demand changes, so a single variation in the initial plan can completely alter the best solution. This is particularly true when the required production capacity gets near to the available. Thus, developing algorithm for solving the MMJIT problem under the hypothesis of constant demand or constant product mix seems useless.
JIT was developed for manual or semi-automated assembly line systems, not for completely automated manufacturing systems. The flexibility of the JIT approach requires a flexible production environment (i.e. the process bottleneck should not be saturated) and this is not an easy condition to be reached in real industries. Consequently, despite the competence of its operations managers, even a big multinational manufacturer may encounter several problems in implementing JIT if a significant part of its supplier is made of small or medium-size enterprises (SMEs), which are naturally more exposed to variability issues. On top of this, differently from MRP – where the algorithm lies within a software and is transparent for users – in JIT the product sequencing is performed by the workforce and managed through the use of simple techniques, such as the heijunka box, the kanban board or other visual management tools, e.g. andons. Thus, any approach to organize JIT production should be easily comprehensible to the workers and should not require neither expert knowledge nor a supercomputer to be applied.
- 1247 reads
