ABSTRACT Currently, the world faces challenges related to pollution and the shortage of raw materials that exploit and damage natural resources. A principal solution to these issues is to adapt sustainable and circular manufacturing, where existing resources and materials are utilized to meet their fullest potential for new demands. To achieve this solution, used and unusable waste electrical and electronic equipment (WEEE) must be efficiently disassembled and adequately processed. Mostly, the WEEEs are selectively disassembled by operators focusing on high‐value parts. Despite numerous existing studies on selective disassembly sequence planning (SDSP), there are still gaps in developing an efficient framework to tackle the multi‐targeted disassembly of diverse products with different structures in SDSP. This work proposes a hyper‐hybrid reinforcement learning model to address the SDSP problem. This proposed model integrates the quantile regression deep Q‐networks and proximal policy optimization network within the actor‐critic framework, which enhances the model's performance, especially when handling complex problems and constraints. Additionally, this model incorporates new parameters to improve stability and convergence rate. Regarding engineering application, the proposed model was tested on 12 products with multiple targets and varying structural complexities. The experimental analysis, comparing the proposed model with state‐of‐the‐art and advanced methods, shows that the proposed model surpasses all existing approaches in terms of optimality while achieving all the given targets. The proposed model can be instrumental in tackling the SDSP problem and achieving artificial intelligence‐based solutions for handling WEEE in sustainable manufacturing processes.
Chand et al. (Wed,) studied this question.