FAULT-TOLERANT FLOW-LINE DESIGN AN EXAMPLE FROM AN AUTOMOTIVE BODY SHOP

Christoph Muller, University of Rhode Island

Abstract

Automotive companies employing highly automated assembly lines in their body shops struggle with the challenge of robots' failures which result either in stoppages of the line or in manual backup of operations. These phenomena tend to impair the systems' productivity as well as the products' quality. Therefore, the consideration of robot failures in the design stage of an assembly line considerably gains in importance. In this thesis we develop an approach to configure a robotic assembly line such that it is more robust against robots' failures. The principle idea is that in the event of a robot failure, working robots perform the tasks of failed robots. The throughput loss in these backup situations mainly depends on the systems' level of redundancy, i.e. the higher the systems' level of redundancy the more robust it is against robot failures. Consequently, we develop an approach to design the line such that the systems' level of redundancy is maximized.

We provide a numerical analysis in which we compare the performance of our approach with that of a traditional robotic assembly line balancing robot. In order to evaluate our approach we use performance measures for the average cycle time and the product quality. In all settings considered we find that our approach performs better as the benchmark with regard to both evaluation criteria. The best performance with regard to the cycle time is obtained for a low to medium number of welding guns per group of weld spots and a relatively low mean time between failure (MTBF). Our approach allows for a reduction of the cycle time of 2 % to 4 % compared to the benchmark. When considering the product quality we get the best performance for a medium to high number of welding guns per group of spots for both values of the MTBF that have been considered. The number of groups of spots that have to be backed up manually could be reduced by approximately 60 %. From this we derive that our approach is best suited for a manufacturing environment which is characterized by frequent but short robot failures and a medium number of welding guns per group of spots.