Bayesian decision-making for industrial production facilities and processing

Abstract

Decision on a strategy for effective predictive Reliability, Availability, Maintainability and Safety (RAMS), by the application of Bayesian networks, while ensuring a better preserving of the operators and installation safety in its entirety. A Bayesian network is an acyclic directed graph where nodes represent discrete random variables value (True, False), and the links influences between the variables or conditional dependencies. Relations between variables are deterministic or probabilistic. In a context of risk management, the causal relationships between different events (cause-effect) that can save any installation dysfunction should be taken into account, integrating the conditional probabilities, based on the opinions of experts' field and on the data mining. Bayesian Networks have become a tool for uncertain reasoning, monitoring tasks such as diagnosis, prediction, and decision making. This makes Bayesian networks a subject of research of artificial intelligence. The processing of data through inference allows us to analyze up-and-down and enrich the basis of feedback through the acquisition of observations (evidence). In this study we present the contribution of Bayesian networks to production and processing of natural gas and an application example will be given for a component (boiler) of the liquefied natural gas complex GL4z industrial facility located in Arzew, western Algeria.