C. ARGIOLAS*, A. CARBONARI**, F. MELIS*, E. QUAQUERO*
(*)
Università di Cagliari, DICAAR Department, Cagliari, Italy
Università Politecnica delle Marche, DICEA Department, "Building Construction and Automation"
Research Team, Ancona, Italy
(**)
argiocar@unica.it
filippo.melis@unica.it
equaquero@unica.it
alessandro.carbonari@univpm.it
Integration of intelligent models and sensing to prevent hardly predictable hazards
in construction sites
Keywords: Pervasive sensors, health and safety, Bayesian networks, construction sites.
This article reports on the actual state of an ongoing research project which is aimed at implementing
advanced probabilistic models for real time identification of hardly predictable hazard scenarios in building
sites.
Safety is currently managed, first, by means of safety planning in order to identify potential risks, then, by
supervising the correct application of the mitigation actions prescribed.
As any programmatic actions cannot deal with the unpredictable nature of many risk dynamics, we tried to
survey on how the current approach for safety management in the construction industry could be improved.
The past research step has already led to a Bayesian network [1] for safety management based on elicitation
of subjective knowledge [2]. The built network can be intended as a reliable knowledge map about accident
dynamics, highlighting the detectability level of the hazardous events [3]. However, the built network shows
that a relevant part of occurrences fall in the category of “hardly predictable hazard”, relatively to those
scenarios which are not explicitly dealt through regular supervision of compliance to safety regulations.
This paper shows the second result of our research project, which focused on the development of
elementary fragments of the overall net, each relative to the occurrence of a possible “hardly predictable
scenario”. Experts’ contribution cannot be helpful to this aim, because in their carrier they hardly incurred in
many of such situations. Instead, we used an accurate source of information, “legal cases”, which
suggested the categories of “hidden hazard scenario” with high frequency of occurrence. Neglecting all the
cases linked to the “total breaking of rules”, which can be dealt through regular surveillance, the most
frequent hidden hazard scenarios are linked to operator’s negligence, abnormal behavior: e.g. uncontrolled
removal of scaffolding’s elements, unprotected openings, improper use of PPE, etc. Each pattern determined
by legal cases has been formalized through a fragment of a network.
Finally, all the fragments will be integrated in a comprehensive intelligent tool for hardly predictable
hazardous prevention. The sensor setup needed to interface to the intelligent models and feed them with
inputs about the current state of the context is discussed, too.
References
1. Pearl J. “Probabilistic reasoning in intelligent systems: networks of plausible inference”. Morgan Kaufmann
2. Vose D. “Risk Analysis. A quantitative guide.” John Wiley & Son LTD
3. Argiolas C., Carbonari A., Melis F., Quaquero E. “A Bayesian Model for real-time safety management in construction sites”.
ISARC Conference 26-29 June 2012, Eindhoven, Netherlands.