Modeling of Technical, Human and Organizational Factors and Barriers in Marine Systems Failure Risk: Modeling of Stability Operations on a Semi-Submersible Unit with the use of Bayesian Belief Networks

Offshore operations in the relatively harsh conditions on the Norwegian Continental Shelf requires a strict safety focus. In order to conduct safe operations, a range of risk analysis tools are needed to monitor the risk level. In recent years there have been a great focus on avoiding hydrocarbon leaks, but in the last 25 years there have not been any in-depth work focusing on marine systems in general, and stability in particular. It is found that most risk analyses treats stability in a superficial way. Furthermore, it is found that most accidents results from inadequate operational safety, and that human and organizational factors often causes these accidents. The objective of this thesis is therefore to develop a model that can be used to analyze the risk involved with technical, human and organizational factors in stability operations.Due to the limited amount of work that have been done in this area, this thesis starts out by defining and explaining the basic concepts of stability of semi-submersibles, and the systems that are installed to perform stability operations. Further, the theory of bayesian belief networks (BBN) is explained. BBN is a method that can be used to analyze risk, and it is particularly suited for handling uncertainty, non-deterministic and non-sequential relationships.A thorough review of technical, human and organizational factors and the interaction between these factors is done. It is found that technical factors often works on the unifinality side of the scale, meaning that there is one way for these factors to perform their tasks. Organizational factors on the other hand are on the equifinality side of the scale, meaning that there are many ways for these factors to perform their work and still yield the intended output. Human factors, are somewhere in between, and acts as the link between the organization and the technical systems. The unifinal properties of technical factors means that they are well suited for traditional risk analysis tools such as fault and event trees. These tools can, however, not be used to analyze the risks involved with organizational factors, due to their equifinal properties. That is why it is suggested to use BBN to develop a risk model for stability operations.A thorough investigation of incidents and accidents has been done to determine the root causes to include in the model. The model has then been built, based on experiences from previous incidents and accidents. In this process it was found that most accidents occurred due to three types of failure:- Failure to conduct normal operations- Failure to respond to abnormal situations- Failure of technical systemsThe model has therefore been built around these three failure types, by focusing on the root causes for why these three failure types occurs. A semi-mechanized algorithm is used to quantify the model. This is used due to the massive workload associated with quantifying a BBN. This algorithm requires expert judgment input about how important each risk influencing factor (RIF) is compared to the other RIFs, and to what extent the RIF will influence its child node. This means that the model to a small degree is based on real data. Because of this it is concluded that the model it not suited to determine a specific risk, but rather to monitor how the risk level is developing over time, and how it develops when certain RIFs change their state. It is also recognized that the model need some more refinement and validation before it can be applied in real situation in the industry