Graph-enabled digital twins for intelligent product lifecycle management: a multi-dimensional approach to design, manufacturing, and supply chain transformation

• Main Author: Lim, Kendrik Yan Hong
• Other Authors: Chen Chun-Hsien
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Product lifecycle management; Digital twin; Industrial knowledge graph; Manufacturing resilience; Mass customization
• Online Access: https://hdl.handle.net/10356/176020
• Institution: Nanyang Technological University

In today’s increasingly dynamic business landscape, industries are faced with global headwinds, intensified competition, and increased volatility driven by factors such as increased demand for tailored products and solutions as well as geopolitical, economic, and climate disruptions. To address these challenges and remain competitive, the product lifecycle management (PLM) paradigm, which allows companies to be in control of their products and services across the lifecycle, has gained attention from industries. This cross-functional approach, enabled by Industry 4.0 information and communication technology (ICT) enablers such as digital twins (DT) and knowledge graphs (KGs), can elevate existing capabilities for customizability and manufacturing resilience. This thesis explores the role of DT and KG in enhancing PLM capabilities, integrating multiple lifecycle phases with emphasis on product design, manufacturing, and supply chain management (SCM). The adoption of these transformative technologies has the potential to enhance existing mass customization strategies such as the product family (PF) approach and Smart Product-Service Systems (Smart PSS), paving the way for intelligent decision support systems. Applicable to a wide range of industries, the use of DT systems with KG as a computational driver can support stakeholders in improving operational efficiencies, and responsiveness for disruption management. The main contributions of the thesis can be categorized into five research areas highlighted below: 1) Establishing an environment-based context-aware DT system to enhance PF design and optimization. To overcome the challenges of identifying optimal product configurations in the planning phase and overcoming user requirement changes in the usage phase, a context-aware DT system incorporating real-world environment aspects is proposed to aid PF reconfiguration and redesign. A novel benchmark and interacting mechanism automatically identifies feasible PF modules, reducing the need for costly and bias-prone expert recommendations, as demonstrated by a case study on tower crane planning and deployment. 2) Developing DT-as-a-Service to advance service-oriented digital manufacturing. Based on the flexibility and versatility of DT in manufacturing processes, these systems are pivotal towards customized production systems. A four-tier technology stack is proposed to construct DT systems using a modular approach and leverages advanced computational methodologies and tools such as extended reality (XR) to support a wider range of shop floor manufacturing processes. Here, DT systems are used as a fundamental to enable Smart PSS and circular economy paradigms to drive sustainability and customizability efforts. These multi-faceted capabilities are featured in two use cases, additive manufacturing, and gearbox assembly line. 3) Enabling causal inference for maintenance operations using cognitive digital twins (CDTs). The intricate nature of large production systems results in challenges pertaining to the root cause identification of product defects. As DT assets typically operate independently and lack cross-domain knowledge-sharing functionalities, a CDT can integrate multiple DT assets and different shop floor processes to generate optimal solutions. Leveraging industrial knowledge graphs (iKGs) for reasoning, feasible solutions can be generated to holistically enhance shop floor productivity. A print packaging manufacturing line is featured to demonstrate its capability. 4) Designing a hybrid supply and production (S&P) DT system for disruption management in multi-echelon networks. The reliance of many companies on multi-echelon supply networks to optimize and streamline downstream distribution operations emphasizes the pivotal role of hybrid S&P facilities. However, these bottleneck facilities are typically vulnerable to demand fluctuations, and existing DT systems usually operate in isolated domains without consideration of SC and manufacturing aspects. Through the proposed S&P DT system, demand disruptions can be mitigated through resilience evaluation, SC replanning, and shop floor rescheduling functionalities, and is demonstrated in a consumer packaged goods (CPG) industrial case study. 5) Building a graph embedding-based mitigation decision support system (GEM-DSS) for manufacturing resilience. As SC disruptions present risks to business operations and end-user satisfaction, manufacturing resilience is a core factor in Make-to-Order (MTO) strategies. Here, two challenges are identified as knowledge incompleteness and the need for automatic solution recommendation. To tackle these issues, an attention-based graph consistently-attributed graph embedding (ACAGE) model is designed to predict missing data relations via link prediction to increase database accuracy. Two computational pipelines are also developed to derive feasible mitigation strategies for alternative supplier selection and material substitution, and the model's effectiveness is rigorously tested in an automotive case study. The significance of the research lies in the development of novel DT technologies to facilitate customizability paradigms across PLM phases, and the exploratory research of integrating iKGs in DT systems for improved decision-making. The research is organized into eight chapters, supported by six industrial case studies to highlight its practicability. Through the valuable insights derived from the findings, the author hopes that the research will offer useful guidance to both industry and academia towards implementing mass customization strategies