The effects of process time variability on a multiple line, dual stage production system with Kanbans
In an ideal world, the just-in-time (JIT) production system holds zero in-process inventory so that production is literally just in time. Such a system is possible only when process times at each stage are perfectly balanced. However, in a real production setting, sources of uncertainty such as the variability in operator performance and unequal distribution of task times present obstacles to the ideal case.
A flexible simulation model was developed to mimic a real-world, multiple line, dual stage production system with Kanbans. The system is based on the manual assembly process of a major product of a medical device manufacturer located in San Antonio, TX. The production line consists of three parallel subassembly stations that feed into a single final assembly station. Subassembly Work-in-Process (WIP) is controlled and moved about via Kanban bins with fixed number of parts.
The effects of process-time variability on the performance of the system are studied to help the manager identify what type of configuration to implement in the current environment and potential areas of improvement. Factors studied included Coefficient of Variation (COV) of subassembly process times, COV of final assembly process times, line balancing configuration, and various buffer levels. Performance measures included output, utilization and cost.
The results of the simulation experiments support the hypothesis that increasing variability in process times has an adverse effect on build rate. Out of the three configurations tested, the base case consisting of a low-high distribution of process times performed the best within an environment of process time variability. The results also showed that for the configurations studied, lowering the WIP level can reduce cost without compromising output levels.
The model is a valuable tool for managers to assess the effects of changes in various parameters of the production system. In the near future, the model will be used by the company to help quantify the benefits of adding automation equipment to the line. For low to medium volume production facilities, the model can be used by managers looking to implement a Kanban pull system for the first time on a simple production process. For the academic community, this work helps to conceptualize one way in which a Kanban pull system can be constructed in Arena. It also serves as a good base model for comparison against similar manufacturing systems. Lastly, it has opened up the door for other researchers in Lean Manufacturing to build upon.