Forensic engineering data for future infrastructure success
ASCE defines forensic engineering as the "application of engineering principles to the investigation of failures or other performance problems". While notoriety is often gained for the investigations and case studies following large catastrophic failures, smaller scale investigations using these same principles occur almost daily on construction projects in the form of non-compliance, incident and near-miss reports. However, these data are often rendered inaccessible upon project completion, and the learning opportunity to the wider community wasted, only remaining as tacit or experiential knowledge.
Additionally, these data are predominantly captured in written prose, an unformatted style, therefore analysis is hindered by the lack of techniques to deal with data of this type. This results in restricting learning to fewer significant cases to avoid inundating people with constant minor alerts or updates.
In considering how these data may be captured, understood and learning extracted, my work is sectioned into two parts.
The first explores, using semi-structured interviews and thematic analysis, the notion of defining failure within construction projects and the existing ways in which forensic engineering data are captured and lessons learnt disseminated. The aim of this part was to contexualise the problem and generate relevant assumptions about the data in order to intelligently overcome barriers to learning from forensic engineering data. While this part of the research is notionally complete, additional insights can always be gleaned from further experience within industry.
The second part of my PhD will determine how informatics techniques can be leveraged to implement learning from everyday forensic engineering by presenting a framework that uses recent developments in both the data analytic and natural language processing fields. My research will demonstrate a hybrid natural language processing and knowledge discovery framework to transform everyday forensic engineering data into a valuable resource exploitable for learning.