Practical inventory models with the warm-up process

Abstract

As the global population continues to grow, there is an increasing need to enhance the efficiency of production processes. On one hand, manufacturing processes face numerous challenges; on the other hand, various machines in the production line require an initial warm-up phase, which intersects with the fields of operations research and optimization. This dissertation explores the introduction of several concepts along with the warm-up process into the manufacturing workflow. It also addresses a range of issues associated with the warm-up in manufacturing, proposing solutions to these challenges. It tackles common problems in the production line, such as shortages, the environmental impact of carbon emissions, and the production of faulty items. The work at hand employs a diverse set of approaches, from mathematical solutions like the application of the Hessian matrix to the implementation of Karush-Kuhn-Tucker conditions. A variety of methodologies have been applied, ranging from analytical approaches to metaheuristics and innovative deep reinforcement learning techniques. The outcomes of this thesis have resulted in three published papers, with two additional works finished. The publications explore the effect of warm-up process in sustainable EPQ model, the effect of machine downtime on warm-up process, presence of shortage and faulty products with warm-up, machine downtime effect along with shortage on warm-up, and finally multi-product lot scheduling problem with warm-up process. The findings can be regarded as determination of optimal total cost for the system which provides higher revenue for corporations. In case of three published papers, this is done due to analytical approach and mathematical framework, in other words, a closed-form solution represents the whole structure. The solution methodology highlights key concepts, such as shortages and environmental regulations, by comparing results that show how the additional cost of carbon policies and the system’s ability to handle shortages contribute to lower overall costs. In cases involving rework and scrap, rework is shown to incur less cost. Finally, the multi-agent reinforcement learning effectively tackled the stochastic nature of metaheuristic algorithms in fine-tuning the control parameters. Altogether, each paper presents a specific direction within this thesis, and collectively, these provide practical insights for decision-makers in the industry.