Enhancing operations for 3D precast construction: a hierarchical planning strategy

Abstract

Singapore has experienced significant advancements in precast technology since the 1980s, when it first adopted precast components, effectively enhancing productivity, quality, and environmental friendliness. Benefiting from the trend of construction industrialization, a precast factory now boasts state-of-the-art machinery, high levels of automation, and advanced planning and scheduling software. However, these modern approaches are primarily designed for two-dimensional (2D) components, whereas the production of three-dimensional (3D) modular units, such as Prefabricated Bathroom Units (PBUs), still involves a fit-out process characterized by the fixed-position assembly that heavily relies on human labor and is far from automated. Due to its human-intensive nature and operational complexity, workforce management in this context faces several challenges: (1) uncertainty and variability resulting from human factors, (2) increased complexity due to multi-skilling strategy, and (3) the dynamic production environment. As a result, the industry still heavily relies on human experience for workforce management and lacks a systematic decision-making tool. Therefore, the motivation of this thesis is to establish an intelligent and reliable approach that can enhance 3D precast construction with fit-out operations, integrating workforce planning and scheduling through hierarchical planning strategy. To achieve the overall research objectives, three major research works are systematically proposed to enhance workforce management. This begins with long-range strategic planning that considers organizational market conditions, followed by medium-range tactical workforce scheduling focusing on resource utilization, and finally, short-range operational workforce dispatching that addresses the dynamic production environment and unpredictable disturbance events. The details are elaborated in research objectives 1 to 3, respectively in Section 1.3. The key findings and contributions are summarized as follows: (1) A mathematical model linking two critical decision variables—workforce hiring and training—and the cycle time was established. Sensitivity analysis based on this model indicates that expanding the skill diversity per worker provides a notably higher return on investment compared to merely increasing the workforce size, highlighting targeted skill development as a more economically effective strategy for optimizing productivity. (2) A two-level hierarchical model incorporating both project and workforce scheduling was developed, capturing individual worker learning effects. The proposed Hybrid Whale Optimization Algorithm (HWOA) demonstrated superior computational performance and effectively visualizes project and workforce scheduling solutions (both internal and external) through automatically generated Gantt charts. (3) A Multi-Agent Deep Reinforcement Learning (MADRL) framework was proposed to achieve dynamic project rescheduling and workforce dispatching, providing yard managers with robust tools to manage operational uncertainties and maintain production stability. Three effective dispatching rules were identified through comparative analysis.