The manufacturing industry is ever-evolving. Circumstances such as fluctuating demand, dynamic supply conditions, and the far-reaching trend of customisation and personalisation, bring the necessary challenges in terms of production planning. In bigger factories, shop floors are laid out to meet the production requirements of today’s world, with back-to-back and parallel production possibilities. The problem manufacturers run into, is that their scheduling tools cannot keep up.
The dynamic market circumstances and highly customised products entail unique, complex production flows. Due to urgent requests from customers and shortages in the supply chain, manufacturers need smart, advanced planning and scheduling tools, that are able to incorporate and react upon those dynamic changes and uncertainty.
Existing production scheduling tools are often rule-based and run by computer power. These tools take long times to solve scheduling problems, e.g., up to an hour for computing a reschedule. This is not suitable for handling the dynamic nature of orders and the flexibility required in make-to-order production environments. This results in tardiness in the entire production process. For example, the quotation process for new projects is slowed down due to this and the production environment is not utilised to its full capacity.
To develop a solution to this, the AMP-subsidised project TIMELY was initiated. The project consortium consists of Limis from Enschede, specialist in production planning software, Hittech Bihca, part of the Hittech Group – a system supplier for high-tech equipment, that manufactures semiconductors or electronic components, and the FIP-AM@UT.
The goal of the project is to design, train, and implement a computationally efficient scheduling demonstrator for hybrid job-shop production systems. The scheduling tool will make use of reinforcement learning, a form of artificial intelligence, for stronger computational power and quicker system responses to changes in the demand or production. It is important that the scheduler can quickly produce a new schedule upon an urgent customer request, while creating robust schedules that account for uncertainty. In turn, this will lead to optimised operations.
