DNV GL Research and Innovation has been working several years to develop the
future risk assessment method for pipelines that can give better, more
correct and real time risk that can be visualized on a GIS system in real
time. The development has resulted in a new tool that is named MARVTM –
Multi-Analytic Risk Visualization tool. The system has been tested out on
several clients globally with very good result.
MARV™ is a risk assessment methodology based on Bayesian networks. MARVTM
uses Bayesian Models Forward Projection Plus Backward Correction. More
importantly, uncertainty in the calculations is quantified too. This
uncertainty is driving the data collection, when the uncertainty is
acceptable, expensive data gathering can stop. The MARVTM tool allows
sensitivity analysis on the results that helps decide between data gathering
and/or mitigation. Bayesian inference is used to update the probability of an
event through observations/monitoring improving the initial
“guess/assumption” through real data. Bayes' equation allows the
probabilities of various outcomes to be updated based upon the relationships
between models and evidences. After inference, the updated probabilities
reflect the new levels of belief of all possible outcomes.
Why is MARV™ different from current risk assessment methods?
• Do not use single values, but a probability density functions
• Links all models no matter origin, complexity, or programing language in
one single frame work.
• Understand where the risk comes from.