Robotic dismantling based on deep learning in WEEE recycling

The proportion of waste of electrical and electronic equipment (WEEE) with components containing hazardous and precious metals will significantly increase in the coming years. Nonetheless, the majority of WEEE is today still processed by large-scale shredder based recycling technologies, which are today the most economically viable option for the majority of WEEE categories. Whereas these technologies have undergone substantial improvements over the past decade, they are in comparison with manual disassembly still characterized by low recycling efficiencies, especially for the recovery of precious metals and the proper processing of hazardous materials. Therefore, there is an outspoken opportunity to automated the dismantling process for mixed WEEE to increase recycling efficiencies in an economically viable manner in support of a transition to a more circular economy.

Due to the high variation in both product models and conditions of WEEE, the required flexibility of automated dismantling systems is exceptionally challenging for these mixed waste streams and goes far beyond reversing state-of-the-art assembly methods adopted in manufacturing. Therefore, this PhD project has the objective to address these challenges by:

1.         Improving product model recognition accuracy by investigating the applicability and efficiency of state-of-the-art computer vision techniques

2.         Optimal path planning for robotic dismantling based on both component and fastener detection

3.         Implementing self-learning principles to enable a gradual expansion of the knowledge based to improve the dismantling efficiency