Reducing Waste in Electronic Components: A Circular Economy Decision Model Using Graph Neural Networks and Reinforcement Learning

This work stems from the need to find a solution to mitigate one of the problems that characterizes most the environmental impact of the modern era: e-waste management. After analyzing its relevance in the context of environmental sustainability, the thesis discusses different types of tests used to study the defects in printed electronic boards and the recovery strategies that can be implemented. The thesis proposes a decision-making model modelled in the form of a Markov Decision Process characterized by uncertainty. The information space of PCBs, consisting of varying numbers of components that may have different defects, was represented through the use of graphs. These main aspects lead to a resolution through the implementation of a simulation algorithm developed in Python based on Reinforcement Learning and the use of Graphical Neural Networks. Through the use of simulated data from the use of historical data and test and recovery strategies, the aim of this model is to identify the optimal test sequence required to understand the defective state of electronic boards and the appropriate subsequent recovery strategy, balancing cost and profit.