Loss of containment insights project

Across the process industries, both on and offshore, a key focus is upon maintaining primary containment. Primary containment refers to keeping flammable and toxic materials inside plant items such as vessels, tanks and pipework. Loss of primary containment can result in fires, explosions and toxic releases with the potential to harm workers, the offsite public and the environment.  

The Buncefield explosion and fires in 2005 resulted from a Loss of Containment (LoC) due to the overfilling of a fuel storage tank. Fortunately, in this case there were no fatalities, but the incident caused an estimated £1Bn of damage to the economy. In many of these events there is the potential for large numbers of fatalities. The incident at a pesticide manufacturing plant in Bhopal, India in 1984 resulted from LoC of a toxic material which affected many people offsite and caused the death of thousands of people both immediately and over the longer term.  

Smaller LoC events and those where no harm occurred are considered precursor events to major accidents. Accordingly, the Health & Safety Executive (HSE) and others investigate and record a wide range of LoC events together with their causation and circumstances.



Our early work set out to explore the feasibility of providing the process industry with better intelligence with which to improve standards of process safety. The approach has been to examine HSE’s own regulatory data within the COIN database, both coded records and free-text. At the same time we have consulted in detail with several stakeholders and have had contact with others at a higher level. We have established that sufficient data exists and there is demand from industry for the type of insights which could be delivered through the next phase of work. We have also explored the value of augmenting HSE’s data with industry’s own datasets and the barriers which need to be overcome to achieve this. We have applied Natural Language Processing (NLP) techniques to the data and early results are very promising. We have also uniquely explored the use of occupational safety (e.g. minor personal injury accident) datasets to augment the directly process safety relevant datasets to greatly increase the number of records and therefore the level of detail we can extract. Engineers and managers in the process industry need better intelligence to improve the management of risk to prevent and mitigate high-consequence outcomes such as fires, explosions and toxic releases. Preventing and mitigating these events reduces the harm suffered by large numbers of both plant workers and the offsite public who can be affected due to the wide hazard range resulting from such catastrophic incidents.

Our next focus is aimed at accelerating key activities to provide better tangible outcomes to demonstrate the value which can be added to industry. We are seeking industry collaboration to augment HSE’s data set with to further develop the possible insights.

Get involved

Seeking industry partners to collaborate with in data sharing, our work with Ohalo enables secure anonymised data sharing possibilities.

Industry drivers

LoC events continue to occur worldwide in both emerging economies and developed countries. Whilst these events happen much less frequently than personal injury accidents such as falls from height, when they do occur the consequences are often catastrophic.  

Such catastrophic events threaten the business continuity of individual sites and organisations. They can also have unwanted consequences for whole industry sectors, since the public will only tolerate the presence of hazardous industries if they are confident the associated risks are well managed.  

The need for good asset management techniques at sites with key infrastructure necessary for the production or provision of public utilities is well established. LoC events threaten the integrity of multi-million pound asset infrastructure, even if they occur on lower-value adjacent plant.  

Efforts to decarbonise energy supply and other industries means that the risk profile of the process industries is likely to change over time. New fuels, feedstocks and processes are being deployed which have different features to those which are long established.  

There are therefore sound business reasons now, and in the future, to understand, prevent and mitigate the potential for LoC events in the process industries.


Aims and objectives

The purpose of this work is to use LoC datasets to develop user-led insights into the causes and circumstances of these events across the process industry. This aims to provide sufficiently detailed analysis to intelligently target interventions that will reduce future LoC events globally and/or mitigate their consequences. The intention is to start by extracting insight from HSE’s existing regulatory intelligence datasets and then collaborate with other public and private sector organisations around the world to augment the data.  

Experience from other initiatives suggests that imposing detailed new data collection requirements on ‘time poor’ end users is likely to result in failure. Instead this project aims to use data analysis techniques which can glean insight from unstructured data (e.g. ‘free text’) and those with data quality issues or created for other purposes.

Key benefits

Collating, analysing and sharing detailed insight about how LoC events occur across the process industry will enable better control of risk which will:  

  • Protect employees and the off-site public
  • Maintain and enhance reputation and consent from stakeholders and regulators
  • Protect business, project and process continuity
  • Safeguard key valuable assets 

In later phases of the project the intention is to develop electronic resources which will allow end-users to create their own detailed queries on the underlying dataset. In this way insight can be tailored to specific applications since the process industries incorporate a very diverse range of sub-sectors, processes and technologies.