Intelligent e-wastes processing for effective end-of-life management

The research presented in this thesis introduces a novel framework to improve the value recovery from e-wastes at their end-of-life stage, alongside with actively controlling the recycling process for enhancing material retrieval from e-wastes.

Material recovery from e-wastes became intriguing subject due to high demand, surge in raw materials’ prices, and recent shortages that affected production lines. Improving material concentration prior to recycling, as the key matter, has already been studied in context of robotic disassembly and disassembly path planning. However, extremely limited attention has been paid to find the optimum end-of-life operation according to the specific characteristics of e-waste products. The aim of this research is to investigate the effectiveness of robotic disassembly and other previously anticipated approach towards material concentration prior to recycling and offer a reliable automated solution in contrast that address future industrial, economic, and environmental concerns.

Hence, this dissertation proposed framework for an intelligent decision support system (DSS) based on defined characteristics of visually recognising e-waste products. It also introduced an innovative destructive material extraction method as a completive approach to manual and robotic disassembly for material concentration. These carefully tailored steps provide a complete picture of the application of projected framework that advantage recycling sector by automating their decision making and fulfil extraction process, as well as contribute to the knowledge of intelligent automation in manufacturing, remanufacturing, and circular economy.

The research starts by reviewing the recent progress in robotic disassembly and material extraction of end-of-life products to construct a trustworthy model for intelligent end-of-life management of e-wastes following by a brief analysis of the benefits and challenges for the automated and advanced material extraction from e-wastes.

Developed visual recognition of e-waste products prior to Fuzzy decision making, consist of specific convolutional neural network (CNN) that has been trained and fine-tuned to seamlessly recognised sample products with high precision. A real-time DSS is also established to validate feasible paths for each e-waste and find the optimum end-of-life route based on pre-defined characteristic of individual e-wastes such as initial price, material, weight etc. Another important aspect of this research is developing an innovative material extraction method that integrates versatility of a 6-axis robotic arm with advanced visual semantic segmentation to find the extraction target point on e-wastes and utilising hole saw as the main extraction tool.

The result in form of proof of concept, indicates that manual and inefficient labour-oriented perception, separation, and decision making can be replaced by developed intelligent recognition and DSS. In relation to material extraction and concentration, robotic material extraction has a profound influence by improving the performance through focusing on specific selected targets.

This novel framework can be adapted for similar material recovery from other type of end-of-life products. The developed DSS is flexible to any modification including adding or removing different recovery operations. In addition, indicated framework works not only in laboratory scale but with further tests can also be implemented and scaled up for industrial operation in recycling plants.