A current trend in the construction industry is that the work is more often done by multidisciplinary and geographically dispersed teams. However, members of distributed teams may have difficulty establishing a shared context (Pallot 2011). The absence of shared context impedes the reaching of mutual understanding, which may significantly impact collaboration effectiveness and efficiency (Pallot et al. 2010; Clark and Brennan, 1991). An increasingly utilized method to facilitate common understanding between members of a collaborative team is the use ontologies. By definition, ontology is an explicit and formal specification of a conceptualisation of a domain of interest (Gruber 1993). Furthermore, ontology is defined as a controlled vocabulary that describes objects and the relations between them in a formal way (Berners-Lee et al. 2001). Finally, ontologies, provide effective machine-to-machine communication capabilities enabling computational entities and services to have a common set of concepts and vocabularies for representing knowledge about a domain of interest (Wang et al. 2002).
In BIMProve, all data related to safety risks is modelled with a specific risk data management ontology. The ontology aims at formally representing risk-specific concepts and their relationships, and metadata that enables the system to better understand and reason about the structure and purpose of the data. Moreover, it constitutes a common data representation for the heterogeneous group of stakeholders participating in the safety risk management on the construction site. The resulting risk ontology will be developed iteratively, which means that it evolves during the course of this project as the requirements for the information model become more and more mature and specific. The first version of the risk ontology is represented in the Figure 1. As can be seen, some class definitions of the ontology are still incomplete. Hence, for example the class “RiskZone” contains no properties at this stage.
Figure 1: First version of the risk ontology
The first version of the risk ontology is defined based on the concepts and terms that are currently identified in BIMProve as essential for risk management. The central class in the ontology is Risk that has certain data properties such as name and description. In addition, the Risk class is associated with a risk level that is divided into several subclasses representing different risk levels. The class Risk can also be linked with different risk types (i.e. fire and fall from heights) and risk zones. In addition, instances of the class Risk can also have images that represent characteristics of a risk.
The objective of the class Image is to capture the knowledge extracted from pictures taken from a construction site. It holds two data properties that are URL (pointing to the storage location of the image) and ID of the image. The Image class is associated with the RiskRelatedObject class that models the concepts that are essential for the above described “Automated Detection of Safety Factors and Risks” service. In more detail, the class defines the information that can be automatically detected from the pictures by the utilized deep learning based computer vision solution. The RiskRelatedObject class has three subclasses that are SafetyNet, Garbage and Barrier. These subclasses represents concrete physical objects that are automatically detected from images. The xMin, xMax, yMin and yMax data properties capture the anchor box where a detected object is located in an image. Furthermore, the data property hasConfidence represents a confidence score that is the probability that a defined anchor box contains the detected object.
As the maturity level of the above described risk ontology gradually increases over the course of the project, it will be published on the GitHub community platform to improve the visibility and adoption of the ontology.
Do you want to know more? Download our deliverable from here and share your opinion with us through our LinkedIn or our Twitter communities!
References:
Berners-Lee, T., Hendler, J. & Lassila, O. The semantic web. Scientific American 2001, Vol. 284, No. 5, pp. 28-37.
Clark, H.H. & Brennan, S.E. Grounding in communication. Perspectives on so-cially shared cognition 1991, Vol. 13, No. 1991, pp. 127-149.
Gruber T. A translation approach to portable ontologies. Knowledge Acquisition 1993, Vol. 5, no. 2, 199–220.
Pallot, M.A. Collaborative distance: investigating issues related to distance fac-tors affecting collaboration performance, 2011, PhD Thesis, University of Nottingham.
Pallot, M., Bergmann, U., Kühnle, H., Pawar, K.S. & Riedel, J.C. Collaborative working environments: distance factors affecting collaboration. Proceedings of the 16th International Conference on Concurrent Enterprising, ICE’2010, 2010.
Wang, X., Dong, J.S., Chin, C., Hettiarachchi, S.R. & Zhang, D. Semantic space: An infrastructure for smart spaces. Computing 2002, Vol. 1, No. 2, pp. 67-74.