Loss of containment insights phase one outputs

Matt Clay


This use case is aimed at developing insights to help the onshore process industries prevent and mitigate major accidents arising from Loss of Containment (LoC) events. Such events give rise to the risk of fires, explosions and toxic releases and can affect the offsite public as well as workers. These events can occur within developed and emerging economies. 

Aims and objectives


Key findings

Analysis of the coded records has been carried out and these were published at the IChemE Hazards Process Safety Conference in May 2019 at Birmingham. We have also submitted a more detailed summary of the work to date to the Process Safety and Environmental Protection (PSEP) Journal, which will be published in 2020.The early results of using Apache OpenNLP on a manually anonymised small dataset show great promise. 



The following are the priority learning and apply across other use cases:  

  • We need to have a clear route to timely access to updated COIN data 

  • We need to streamline the approach to deploying software in a secure environment to work on real data 

  • Replica datasets need to be created to retain data well beyond seven years. This may necessitate some degree of anonymisation. 

  • Data sharing arrangements need to continue to be developed. 

  • We need to address automated extraction of ‘causation’ from free text which looks at the totality of free text and comes to a reliable conclusion in a similar way to a subject matter expert would. 

  • We need to convince organisations in the private sector to volunteer data to augment the HSE data by addressing their concerns. 

  • Data quality audit would be welcome.