Automated knowledge discovery and data-driven simulation model generation of construction operations

Computer simulation models help construction engineers evaluate different strategies when planning field operations. Construction jobsites are inherently dynamic and unstructured, and thus developing simulation models that properly represent resource operations and interactions requires meticulous input data modeling. Therefore, unlike existing simulation modeling techniques that mainly target long-term planning and close to steady-state scenarios, a realistic construction simulation model reliable enough for short-term planning and control must be built using factual data obtained from ongoing processes of the real system. This paper presents the latest findings of authors' work in designing an integrated data-driven simulation framework that employs a distributed network of sensors to collect multi-modal data from construction equipment activities. Collected data are fused to create metadata structures and data mining methods are then applied to extract key parameters and discover contextual knowledge necessary to create or refine data-driven simulation models that represent the latest conditions on the ground.