Bayesian Network Modeling of Corrosion – Theory and Applications

Narasi Sridhar, MC Consult LLC (retired from DNV), USA

Abstract

Bayesian network modeling (BN) is a class of artificial intelligence and machine learning techniques that relies on building probabilistic cause-effect relationships between various factors affecting an event e.g., corrosion of a system. The uncertainties with respect to data are included in the Bayesian probabilities. The uncertainties in models (cause-effect relationships) are included in conditional probabilities. BN’s are useful for modeling corrosion of engineering systems because there are many uncertainties (and missing information) in these systems and the system performance is dependent on a combination of design and operational factors, including human factors. This talk is aimed at both researchers and practitioners. The talk is divided into three parts: Part 1 will discuss the basics of BN and the various types of BN’s (dynamic BN, hybrid BN, Learning BN, etc.). Part 2 will discuss the application of BN for three industrial problems to illustrate the general approach: seawater corrosion of corrosion resistant alloys, stress corrosion cracking of onshore pipelines, and gaseous hydrogen embrittlement of pipeline components. Part 3 will discuss how BN can be used in research to identify causal connections. The talk will conclude with future directions of BN models.

You can register (free of charge): here. After registering, you will receive a confirmation e-mail containing information about joining the webinar.

This is the third webinar of Season 5 (link to previous webinars).