Bayesian Network Modeling of Industrial Systems

Abstract Typically, machine learning models rely on a lot of data to develop correlations. However, many industrial systems do not have a lot of data, or if they do they are of the wrong kind. Bayesian network models rely on our knowledge developed through painstaking research. In this paper, the applications of Bayesian network to a variety of industrial systems undergoing corrosion is described. The fundamental aspect of this paper is the use of mechanistic understanding to develop predictive models. The approach is illustrated for upstream oil and gas, refinery, geothermal systems, and CO2 pipelines.