Concept for Human-Machine Interfaces for Resilient Data Extraction from Digital Twins

With the rising digitalisation of the construction industry, digital twins evolve as a tool for data management. The accumulation of data in a digital twin over the lifecycle of an infrastructure building yields extremely large datasets. This creates the need for an intuitive interface to extract data from the digital twin. The fast, reliable, and user-friendly extraction of specific information from this data source will be key for safe and resilient operation of infrastructure in the future. Chat bots leveraging deep learning techniques for natural language processing (NLP) have evolved quickly as human-machine interfaces (HMI). This work picks up the idea of using NLP in an HMI and investigates the data processing that is necessary to enable data extraction from a digital twin. The concept is based on a natural language inquiry, which is preprocessed to extract the intentions of the information request. With the intentions, the context for a question answering (QA) task is chosen, which will be used to extract the answer from the digital twin by a deep learning model. Furthermore, this work will discuss the challenges and opportunities which the described concept would face upon implementation.

Keywords: Digital twin, Construction, HMI, ChatBot, Deep learning, Knowledge management, AI, Project management, Data structure.

Download PDF