Evaluation of stochastic bounds on the remaining completion time of products in a buffered sequential workflow

Agile production systems face major issues in satisfying fickle market needs in highly demand-driven industry sectors, such as electronics and mechatronics. In this context, the time needed to complete the production of an item tends to be highly variable, and online estimation of the remaining completion time may suffer the lack of adequate sensor data, especially in existing manufacturing systems. To solve this issue, we propose a new analytical technique for the evaluation of an upper and a lower stochastic bound on the remaining completion time of a product, considering an assembly line made of sequential workstations with transfer blocking and buffer capacity. The approach notably encompasses service times with non-Markovian distribution, and avoids the limitation of existing works requiring the system to be at steady state at the inspection time. The technique is experimented on a case study and validated through simulation, providing an empirical analysis of its complexity.