Data-driven modeling and performance analysis of production system : A case study of the auto lever production line

Abstract

In this paper, we suggest data preprocessing method to model production system and analyze a performance of a serial line having identical machines that work in turn. A four type of data for modeling are defined and several data preprocessing algorithms based on defined data have been developed to calculate cycletime, downtime and buffer capacity. it is shown that there are the five kind of different production type which has different cycle-time distribution and production type 3 occupies more than half of total productivity. we confirmed that this line can also be applied to modeling based on aggregation method in production system engineering by analyzing production rate, blockage and starvation through the simulation. Finally, an asynchronous exponential model is used for modeling of production system, and an estimation of throughput is obtained from model. The difference with actual TP for product type 3 was about 7.74% which is typically acceptable error. For a bottleneck identification using arrow method with calculated parameter, the bottleneck machine of production type 3 is identified as m18. Finally, two kinds of improvements method are suggested through the bottleneck mitigation.