Data-driven Downtime Bottleneck Detection in Open Flow Lines

Bottleneck of a manufacturing system is the resource with the largest sensitivity on the overall throughput. The bottleneck detection is an important problem for manufacturing system improvement. This work proposes a Downtime Bottleneck (DT-BN) detection approach of open flow lines based on the Discrete Event Optimization (DEO) modeling framework using field data. The DEO model enables to identify the machine whose downtime has the largest sensitivity, without calculating the sensitivities of all the machines. The DEO model is a mathematical programming representation of integrated sample-path simulation-optimization problem, i.e., the structure of simulation model is embedded with the optimization problem. The Benders decomposition is applied, and a simulation based cut generation approach is used, which reduces the computational effort without any approximation. Numerical results have shown that the proposed approach performs both effectively and efficiently. Furthermore, the effectiveness can be further improved by gathering a larger set of data, as the convergence of this approach is both proved theoretically in previous research and validated numerically in this work.