The Potential for Using Big Data Analytics to Predict Safety Risks by Analysing Rail Accidents

We are currently going through an unprecedented era of digital transformation. This change includes everything from the Internet of Things (IoT) through Big Data (BD) to new analytical approaches (Analytics) to analysing business and personal needs. The rail industry in particular is a focus for digital transformation through passenger and infrastructure related initiatives. Digital transformation will fundamentally change the way the industry works particularly in regard to risk assessment and safety. With modern powerful computing and the explosion in data availability, from ever expanding sources, there should be opportunities to use a BD approach to flag up high risk scenarios on the railway before accidents occur. In this paper, an understanding of BD as it applies to railway safety management has been developed in terms of the 5V (Volume, Velocity, Variety, Veracity and Value) model. An enterprise data taxonomy (EDT) has also been suggested as a way of bounding safety data items. Three major accidents have been reviewed and assessed to understand what the high level causation was due to. Each of the causes was then linked, using the EDT, to sources of data that may have highlighted the risk of the hazard together with its potential to propagate into a railway accident. The data sources have been evaluated according to the 5V model and assessed for their respective ‘BDness’ to provide a score that can then be interpreted as the potential for big data analytics (BDA). The analysis of the accidents has shown that BD could potentially help in mitigating accidents where the causes are systematic and complex in nature. An enhanced ELBowTie methodology has been introduced to provide a mechanism for feeding BD into safety risk assessments. This methodology will also provide a means for linking real time data updates into the ELBowTie thus enabling a risk dashboard to be envisaged. The deliverables from the research presented in this paper therefore lead to a greater understanding of BD and the methods (analytics) that are needed to make improvements to railway safety.