Efficient and sustainable scheduling strategies for smart manufacturing systems

Abstract

Propelled by Industry 4.0 technical advancements, smart manufacturing is increasingly prioritized and implemented across various manufacturing systems. In this context, technology-supported scheduling plays crucial roles in ensuring smooth manufacturing processes, enabling agile responses to orders, and reducing operational costs. Two crucial aspects of this domain should be emphasized. First, from a physical perspective, the introduction of autonomous mobile robots (AMRs) automates laborious material handling tasks, enhancing operational efficiency. However, this also significantly increases the system complexity and necessitates precise production scheduling to accommodate complicated machine-robot interactions. Additionally, analysis of the robot fulfillment process reveals wastes of energy raised due to the mismatching between robot operations and machine processing. Realizing the lack of studies considering energy elimination from the perspective of facilitating operational collaboration, this dissertation (in Chapter 3) investigates an energy-aware robotic job shop scheduling problem. To conquer the complexity induced by machine and robot operations and enhance the collaboration between processing and moving, network-based energy-aware modelling approaches are developed. Computational experiments show their capabilities in reducing carbon emissions and maintaining throughout. Second, the integration of CPS has been instrumental in connecting the physical and digital worlds. Traditional scheduling methods, which typically rely on expert experiences, frequently overlook the complex interplay of various real-world factors, leading to impractical or inefficient schedules. The advent of IoT enables the collection of vast amounts of data from physical systems. It is thus promising to uncover useful patterns from historical data and incorporate such data-driven insights into decision optimization processes to derive efficient schedules that are highly applicable to real-world scenarios. Motivated by scheduling challenges in real-world systems and the lack of studies considering the incorporation of multiple realistic factors on production efficiency into the scheduling process, how the multiple factors during the production process can influence job processing status are explored (see Chapters 4 and 5). Moreover, whether the influences can be captured and utilized to enhance production scheduling is explored. Specifically, the study in Chapter 4 aims to jointly predict the job processing time and processing rate level to facilitate resource allocation and timely reporting of production status. A multi-input modules-supported dual-task learning model is proposed, which achieves good performance by capturing influences from various aspects within the performing sequence and leveraging the synergy between dual learning tasks. The study in Chapter 5 further develops context-based scheduling method, which integrates the prediction of context-based job processing rate (CBPR) under varying execution scenarios to the optimization process. A CBPR-guided branch-and-price-based scheduling approach is proposed, which can effectively identify promising execution positions for individual jobs so that overall production efficiency is substantially enhanced. To conclude, this research is devoted to developing efficient and sustainable scheduling methods for smart manufacturing systems. The whole work focuses on two main perspectives: (i) coordinating operations in complex robotic production cells to achieve green production, and (ii) deriving data-driven prediction and scheduling optimization methods to timely inform and maximize production efficiency. Important academic and practical insights are generated.