New perspectives at the boundaries of probabilistic risk management

By Claus Myllerup

Known knowns

Risk-based approaches address both the likelihood of something going wrong and the severity of the consequence of it doing so. These two factors together determine the criticality of failure modes, from low risk - neither likely to occur nor terribly momentous if they do - to extreme risk - both likely to happen and catastrophic when they do. Risks are managed and criticality is lowered by reducing the likelihood of a failure by mitigating its consequences or both.

Modern risk management recognises its own weaknesses and works to minimise them. Typically, to determine the likelihood and consequence of failure of a component requires a large amount of statistical data on its lifetime and failure modes. The fourth generation of risk-based maintenance, such as Arivu provided by the Lloyd's Register-owned Knowledge Based Management (KBM), uses the embedded risk management systems to harvest these data dynamically, feeding them back into the system to improve the modelling.

Known unknowns

Risk management is largely experienced based, but for new designs, experience can be a fickle guide. Blindly extrapolating outside the envelope of our previous experience can be a risky practice. Many of the larger critical components, such as major rotating machinery and cryogenic units, are unique or tailormade and even where there is experience, there is rarely the abundance of data you need to establish the full statistical foundation for conventional risk management.

At these levels of innovation, a thorough technical assessment of potential failure modes can weave a safety net for new designs. The inherent uncertainties of such modelling can be accommodated in a probabilistic framework using methods such as first order reliability, Monte Carlo simulation and so on. Companies like Lloyds Register Martec have successfully applied such modelling to supplement the relatively scarce statistical data available for innovative design of risk critical components.

Unknown unknowns

Modern energy infrastructure is a triumph of specialisation and synthesis. We assemble our assets from hundreds of systems, each in itself a marvel of technological ingenuity. But exactly this level of specialisation and modularity introduces fault lines in our risk management and safety paradigms. Each system is understood intimately, but individual systems interact with each other in complex, non-intuitive and sometimes surprising ways.

Small wonder then that despite our diligence in applying the knowledge from what we have seen before and regardless of our resourcefulness in anticipating what we haven't, we will always find ourselves, now and again, faced with failure.

When structures, machinery or components fail, our imperative is to make sure it doesn't happen again. Clearly it mustn't happen again on the asset in question, but we must also look to the longer term, capture the failure mode and incorporate it into our risk management protocols. Only then can we be sure it is properly accounted for next time we design and build.

But to ensure a failure doesn't reoccur, we must be confident that we know and understand the relevant causes of that failure. These causes will certainly include an account of physical mechanisms, but just as important to both understanding the problem and ensuring it doesn't reoccur, they should extend to the procedural and human contributions in the network of causes leading the failure.

A mechanical analysis can highlight human and procedural shortcomings, just as an analysis of procedure and where it breaks down can raise mechanical questions and inform analysis of human performance. Many failures lend themselves to one approach or another, but only when all three are represented in the process can we be sure we have captured the entirety of the failure mode. Only then can we ensure that all the options for mitigating the problem in future are covered and the measures for doing so are optimised.

The recent acquisition by the Lloyd's Register group of Lloyd's Register ODS (mechanical failure analysis), Human Engineering (human factors) and Scandpower (risk and safety) brings all these fields under one umbrella, facilitates mutual learning and ensures that each of these companies has immediate access to the expertise of the others so that they may provide a comprehensive picture of what went wrong and a confident assurance that it won't happen again.

Biography

Claus Myllerup is Managing Director of Lloyd's Register ODS. Myllerup has served as Chairman for the American Society of Mechanical Engineer's International Gas Turbine Institute Conference and is an external lecturer at the Technical University of Denmark. He has a PhD in mechanical engineering.