Utilising Bayesian networks to demonstrate the potential consequences of a fuel gas release from an offshore gas-driven turbine

Abstract

This research proposes the application of Bayesian networks in conducting quantitative risk assessment of the integrity of an offshore gas driven turbine, used for electrical power generation. The focus of the research is centred on the potential release of fuel gas from a turbine and the potential consequences that follow the said release, such as fire, explosion and damage to equipment within the electrical generation module. The Bayesian network demonstrates the interactions of potential initial events and failures, hazards, barriers and consequences involved in a fuel gas release. This model allows for quantitative analysis to demonstrate partial verification of the model. The verification of the model is demonstrated in a series of test cases and through sensitivity analysis. Test case 1 demonstrates the effects of individual and combined control system failures within the fuel gas release model; 2 demonstrates the effects of the 100% probability of a gas release on the Bayesian network model, along with the effect of the gas detection system not functioning; and 3 demonstrates the effects of inserting evidence as a consequence and observing the effects on prior nodes.

Development of an Algorithm to Convert Linear Belief Function inputs to Exponential Conditional Probability Functions for Multiple Method Applications

“…This conditional probability of D l is given by the sum of the product of the relative weights of each parent and each linear compatible parental configuration. For further information on the symmetric method and the weighted sum algorithm see the following sources (Das, 2008), (Loughney & Wang, 2017) (Loughney et al 2018). This algorithm was applied to the ER belief degrees and relative weights to complete the large CPTs generated by having a number of parent nodes with five states (belief degrees) into a child node with five states.…”

Applying a Bayesian Network Methodology to an Offshore gas Turbine Driven Power Generator to Demonstrate the Cause and Effect Relationship of the Turbine Running Over-speed and the Associated Switchboard Failures

“…The procedures varying depending on the context of the model and level of data available. The methodology for constructing the BN is adapted from previous research in Loughney & Wang (2017) and Loughney et al (2018) (Fenton & Neil, 2013) (Loughney & Wang, 2017) (Loughney, et al, 2018).…”

“…Gas turbine generator sets, particularly aeroderivative gas turbines, are sensitive to electrical faults that can severely damage the turbine and surrounding equipment. Hence, this new BN demonstrates the potential HC release consequences of an offshore gas turbine running overspeed along with the potential ignition consequences resulting from an electrical overload (Loughney & Wang, 2017) (Loughney, et al, 2018).…”

“…These final consequences were not expanded or demonstrated in detail to keep the complexity of the model as low as possible while achieving valid results. Similarly, Loughney et al (2018) further developed a BN model that focused on gas turbine fuel gas release and the resulting ignition consequences (Loughney, et al, 2018). This research expands on this by constructing a new BN network that combines the HC release scenarios of Loughney & Wang (2017) and the ignition scenarios of Loughney et al (2018).…”

“…Similarly, Loughney et al (2018) further developed a BN model that focused on gas turbine fuel gas release and the resulting ignition consequences (Loughney, et al, 2018). This research expands on this by constructing a new BN network that combines the HC release scenarios of Loughney & Wang (2017) and the ignition scenarios of Loughney et al (2018). Gas turbine generator sets, particularly aeroderivative gas turbines, are sensitive to electrical faults that can severely damage the turbine and surrounding equipment.…”