Development of a Deep Reinforcement Learning Algorithm in a Dynamic Cellular Manufacturing System Considering Order Rejection, Case Study: Stone Paper Factory

In this research, a deep reinforcement learning algorithm is proposed for the cellular manufacturing system problem considering the costs of delay and rejection of orders. Orders with different characteristics including revenue, lead time, delivery date, and delay cost are dynamically entered into the system at different times. Due to the limited capacity of the system, it is not possible to accept all orders and some of them must be rejected at the time of entry to enable timely execution of other orders. A mathematical model with two objectives of maximizing profit and minimizing the number of rejected orders is presented and a deep reinforcement learning algorithm is used to solve this problem. The proposed algorithm is compared with the algorithms available in the literature in different categories of example problems and real problems and its efficiency is proven. The results show a 36.3% advantage in profit and 13.87% in the number of accepted orders. Also, by accepting 1% more orders, the profit decreases by 2.7% on average