RFID-enabled Real-time Advanced Production Planning and Scheduling using Data Mining

Abstract

RFID (Radio Frequency Identification) technology has been widely used in manufacturing companies to support their production decision-makings such as planning and scheduling. Significant benefits have been obtained like real-time data collection, advanced production planning and scheduling (APS), as well as efficient material tracing & tracking. However, these companies are dazed when facing vast amount of RFID data, which could be further processed to obtain some invaluable knowledge for advanced decision-makings. This thesis proposes a holistic RFID-enabled solution for manufacturing companies which are facing typical challenges like paper-based data collection, inefficient planning and scheduling, ineffective work-in-progress (WIP) items visibility and traceability, as well as unsynchronized decision-making procedures. This solution includes several aspects. Firstly, RFID devices are systematically deployed in manufacturing sites (e.g. warehouse and shopfloors) to create an RFID-enabled ubiquitous production environment, where typical resources are converted into smart manufacturing objects (SMOs) which are able to sense and interact with each other. Thus, production logics could be carried out adaptively. Secondly, a real-time production planning and scheduling model is worked out for suiting the RFID-enabled ubiquitous manufacturing environment. This model uses several key concepts like hybrid flow shop scheduling (HFS), real-time job pool, and hierarchical decision-making principle to integrate production planning and scheduling level interactively. A real-time Kanban is proposed to coordinate these two levels. Thus, production decisions achieve a real-time fashion. Thirdly, in order to make full use of the RFID-captured real-time shopfloor production data, a data mining approach is introduced to excavate invaluable information and knowledge for APS decision-makings. Standard operation times (SOTs) and decision rules are mined for this purpose. Fourthly, an RFID-enabled real-time APS model is proposed for production decision-making. The resulting APS model is based on a hierarchical production decision-making principle to formulate planning and scheduling levels. An RFID-event driven mechanism is adopted to integrate these two levels for collaborative decision-making with the data mining approach. An RFID-enabled real-time advanced production planning and scheduling shell (RAPShell) is developed by using the concepts and models proposed in this thesis. Some cutting-edge technologies are implemented within RAPShell such as service-oriented architecture (SOA), Software as a Service (SaaS), and XML-based (re)configuration. A case study from a real-life automotive manufacturer is presented for demonstrating how RAPShell is able to facilitate the production activities and decision-making procedures. Benefits from quantitative and qualitative aspects in this case are summarized and discussed. Some innovative contributions are significant. Firstly, an affordable and systematic RFID deployment scheme is proposed to create an RFID-enabled ubiquitous manufacturing environment. Secondly, an entire data mining approach is worked out for discovering the invaluable information and knowledge from vast amount of RFID production data. Thirdly, an APS model using RFID-event driven and data mining technique is proposed to achieve ultimate APS within the ubiquitous manufacturing. Finally, insights and lessons learnt from this research and implementations are generated as managerial implications which could be referred by both academics and practitioners when contemplating the RFID-enabled solution.