presented a comprehensive set of safety risk drivers addressing the most frequent types of injuries in the construction industry, fall
Path of mobile scaffolds and lifts Class 3 (pathfinding)
5。Discussion
A study on more recent and complete data than what Behm (2005) investigated previously, as well as two other sets of accident
and falling objects, electrocution, unsafe machinery operation, and
cave in and asphyxiation。
There are a noticeable amount of construction safety issues that seem unlikely to be identifiable by analyzing Building Information Models。 Concepts like safety climate do have a confirmed effect, and object oriented risk drivers cannot manifest all the conditions influencing the accidents。 This can make some clients hesitant to spend their time and resources on checking safety risk drivers。
454 H。 Malekitabar et al。 / Safety Science 82 (2016) 445–455
Provided that the drivers introduced in Tables 5–9 are checked, however, Table 3 demonstrates that a high proportion of accidents
-ranging from 20% for machinery-related to 60% for fall accidents- can be identified by an automated model checking engine。 As discussed in the methodology section, the drivers introduced here cover the majority of safety issues since the reviewed accident reports are a good representative of all the accident types。 To indicate how possible it is to check the existence of drivers in a model, this research uses a classification suggested by Solihin and Eastman (2015)。 Rules that fall in lower classes are easier to process than those categorized as class 3, which may require new concepts and algorithms to be developed。 Moreover, drivers entailing the acquisition of external data (class 3b), or technical simulations by specialized software (class 3c), can be processed depending on the interoperability capabilities of the software in use。
A set of risk drivers can always be open to new issues, to address as much hazards as possible, according to the databases and technologies available。 Analyzing more accident reports, both formal and unofficial, will result in a richer set of risk drivers, and hence, an inclusive risk identification output。 Formal reports may lack generality, because many accident cases are never reported, in return, unofficial reports are found not to cover all aspects, for example, structured interviews with the witnesses of the event。 In addition to the accident reports, near misses have to be reported and carefully analyzed, because they will contain precious signs of the attenuate drivers that can be regarded as opportunities to be managed。
Future research has to specify a ranking to the risk drivers。 The results of a model checking process have to indicate which drivers are going to stimulate more serious hazards。 Relevant drivers can- not simply be added to give an overall probability, since they are not usually of the same type。 Some drivers may be scored as a yes or no criteria, others as a single or a range of values。 Some may act in similar directions, but together not be as powerful as another single driver, and some may act in different directions, without any information as to whether they are actually canceling each other out or one will be dominant。 This will require a struc- tured determination of attributes of fuzzy numbers assigned to each driver, to reflect the various conditions and interactions they are subjected to。
6。Conclusion
Most construction accidents are preceded by various types of events or conditions that signal something wrong is going on, and signals from a significant number of them are already present during the design phase, but little information on how they can be effectively captured is provided by the literature。 Building Informa- tion Models can support the detection and interpretation of such signals, as they include almost all the objects and their relations in a project。 Yet, current BIM-based tools check only one or a few rules from safety standards, though it is found that not all acci- dents necessarily happen through violations of standards。