201207 – Hughes – You Can’t Get Good Train Control From Poor Data

The past century has seen major improvements in train control. In the early days of the railways, many safety controls were created only in response to accidents. In the 21st Century, before being able to operate, Australian railways now have to demonstrate that they have reduced risk to train control So Far As Is Reasonably Practicable (“SFAIRP”). However, good risk management is only possible when there is good quality data: past incident and accident data is vital to being able to make improvements in the future.

Whilst there have been enormous advances in information technology in the past few decades, data collection techniques have not kept pace with these changes. It is still not possible to get answers to even simple questions about the safe control of trains: for example the available data do not tell us how many derailments in Australia are ‘high consequence’ events (involving significant damage and / or injury), compared with ‘low consequence’ events.

There are a number of emerging technologies that can help continue to improve the safe control of trains. Without reliable data, however, we have no way of knowing how effective these technologies will really be. Examples are available where lack of data has meant that significant investment decisions for train control have been made based on small data sets, or cherry-picked examples of specific accidents.

Moves are underway in Australia to establish a single repository of all railway safety data. At this stage, however, it is not clear how much data will be collected, how the data will be made available, nor when it will be available. This paper discusses how the availability of data is creating real issues for the safe control of trains, and the further changes that are needed to continue improvements in train control in the future.