Machine learning based adaptive production control for a multi-cell flexible manufacturing system operating in a random environment

An adaptive production control approach is used for controlling a multi-cell FMS with machines subject to failures, operating in a highly changing produce-to-order environment. A probabilistic machine learning procedure is integrated within a two-level Distribution Production Control System (DPCS). This enables the DPCS to adapt itself to large fluctuations in demand as well as to other stochastic factors. An extensive simulation study shows that the proposed adaptive control approach significantly improves the production system performance in terms of a combined measure of throughput and order tardiness. The proposed DPCS can be easily implemented as a real-time DPCS due to its simplicity, modularity and the limited information it requires. The proposed adaptive scheme can be integrated in any parametric production control system.