Design Requirements of an Automated Data-driven Simulation Model Generator for Construction Operations

Most construction projects take place in uncertain operational settings where formulating resource interactions and jobsite complexities prior to commencing the project is not a trivial task. As a result, simulation modeling has a great potential to streamline the decision-making process by providing a convenient means to examine different operational scenarios, perform sensitivity analysis, and determine best combinations of resources to finish a project on time and within the specified budget. However, existing construction simulation systems are mostly suitable for early planning stages as they rely on historical data and human expert judgments to create conceptual simulation models. Such systems often fail to properly link real project data and uncertainty characteristics at the execution phase to key simulation parameters and design variables. As such, the resulting simulation models rapidly become obsolete as they cannot keep up with the real world events evolving over the course of the project. Generating simulation models at the operations level with the goal of providing reliable and timely decision support to project engineers and field personnel requires that process data describing the dynamics and uncertainties of field activities be meticulously captured and integrated into the simulation modeling process. To this end, this paper presents the latest finding of an ongoing research which aims at designing and testing an integrated data-driven simulation framework by collecting and fusing multi-modal process data from construction resources involved in various field activities. A reasoning process comprising of data mining, clustering, and fusion techniques are used to analyze raw data, form metadata structures that reveal spatiotemporal patterns hidden within the data, and generate a computer interpretable layout of the jobsite that can be readily used for simulation modeling. Currently, the authors are validating key data mining and knowledge extraction algorithms using several operational scenarios that include multiple types of construction resources interacting in various simulated settings.