Data-Driven Analysis Framework for Activity Cycle Diagram-Based Simulation Modeling of Construction Operations

摘要

This paper proposes a framework for autonomous data-driven discrete-event simulation (DES) modeling process that transforms operation data to Activity Cycle Diagram (ACD)-based DES models without a priori model network. The proposed framework uses the idea of process mining to discover a real process by extracting knowledge from activity logs, which are the lists of sequences of activities per cycle in terms of resources. The discovered process is then modeled as an ACD which can replicate the underlying activity logs. Activity log plays a key role in this procedure as it determines the set of activities that we can identify and thus govern the analysis scope of the generated simulation model. It is therefore important to design and collect activity log data in a way to capture all the activities for an intended abstraction level. In order to automate the model generation procedure considering this requirement, we employ the notion of activity ontology, which specifies various activities and their relations for a given construction operation at different abstraction levels in a predefined manner. Using the ontology-based framework, we can systemize the procedure of simulation modeling, which entails selecting the abstraction level of simulation model, identifying the list of activities that need to be inferred from data to create activity logs, and determining the type and resolution of data to be collected. The proposed framework is explained and demonstrated using an earthmoving operation example.