LOPA – Validating Human IPLs and Ies

Layer of Protection Analysis (LOPA) is a simplified risk assessment method that provides an order of magnitude estimate of the risk of a potential accident scenario. Humans can be the cause on an accident scenario (the Initiating Event [IE]) or human can serve as an independent protection layer (IPL). In either case, estimating the likelihood of the human error and measuring the human error rate at a site are troublesome tasks within the LOPA framework or QRA framework, which is one of the reasons some companies do not give any credit for a human IPL.

Identifying and sustaining independent protection layers (IPLs) is the heart of LOPA. Each IPL must of course be independent of the initiating event (IE) and the other IPLs and must be capable (big enough, fast enough, strong enough, etc.). But also, each IPL must be 1) maintained or kept in practice/service, 2) validated/proven to provide the probability of failure (PFD) chosen, and 3) documented and audited periodically to substantiate these values.

One type of IPL is a Human IPL. These include preventative steps that may stop a scenario from progressing once it is initiated, but more typically the human IPLs are responses to alerts or alarms or troubling readings and sample results.

This paper continues from the paper in 2010 GCPS and from the approaches outlined in ?Initiating Events and Independent Protection Layers, CCPS, 2011?, and shows the data needed for adequately counting the human in a LOPA (and other risk assessments). The main focus of the paper is on practical means for collecting raw data in a plant setting for substantiating the error rates for the site so that the IE rates and IPLs probabilities of failure can be fairly credited. The discussion covers the training requirements that should be met, proof drills for response to alarms, simulations and tests, frequency of proofs, and of course the effect of human factors on human error rates. Actual plant data and tests are included in the paper to provide the reader with examples of how similar data collection and validation can be set up within their companies. The paper also illustrates how such data can be used for adjusting calculation/prediction of human error rates for new settings, such as plant expansion.