The Kanbans are distributed to the next to last downstream workstation first and slowly moved upstream. Their basis for this procedure is the following intuitive argument. This phenomenon occurs
because at upstream workstations the chances of being blocked are higher, while at down-stream workstations the chances of starving are higher. Another intuitive argument is that when Kanbans are allocated at the down stream workstations, the bottleneck is moved to the upstream workstations. However, when Kanbans are allocated to the upstream work-stations, the bottleneck is moved to the down stream workstations.
3.1. Stochastic models for determining number of Kanbans Stochastic demand models are very complex in nature. Most authors have used Markovian or queuing theories for the optimization models, but simulation is very widely used.
3.1.1. Tandem and cyclic queues. Berkley [4, 5] emphasized the problem of setting the number of Kanbans to achieve a given per formance level. In his papers he determines the base configurations of a system with a minimum number of Kanbans under idea conditions. Once the base configuration having the desired production rate is found, it is used
to establish minimum performance levels for each station. Two experiments are shown to provide sufficient but not necessary conditions to guarantee desired production rates indepen dently of the average station processing times.