Statistical Change Detection for Diagnosis of Buoyancy Element Defects on Moored Floating Vessels Mogens Blanke ∗,∗∗ Shaoji Fang ∗ Roberto Galeazzi ∗∗ Bernt J. Leira ∗∗∗ ∗ Centre for Ships and Ocean Structures, Norwegian University of Science and Technology, NO 7491 Trondheim, Norway ∗∗ Automation and Control, Department of Electrical Engineering, Technical University of Denmark, DK 2800 Kgs Lyngby, Denmark ∗∗∗ Department of Marine Technology, Norwegian University of Science and Technology, NO 7491 Trondheim, Norway email: mb@elektro.dtu.dk, shaoji.fang@ntnu.no,rg@elektro.dtu.dk,bernt.leira@ntnu.no Abstract: Floating platforms with mooring systems are used extensively in off-shore operations. Part of the mooring systems are underwater buoyancy elements that are attached to the mooring lines. Loss or damage of a buoyancy element is invisible but changes the characteristics of the mooring system and alters its ability to provide the necessary responses to withstand loads from weather. Damage of a buoyancy element increases the operation risk and could even cause abortion during an oil-offloading. The objective of this paper is to diagnose the loss of a buoyancy element using diagnostic methods. After residual generation, statistical change detection scheme is derived from mathematical models supported by experimental data. To experimentally verify loss of an underwater buoyancy element, an underwater line breaker is designed to create realistic replication of abrupt faults. The paper analyses the properties of residuals and suggests a dedicated GLRT change detector based on a vector residual. Special attention is paid to threshold selection for non ideal (non-IID) test statistics. Keywords: Change Detection, Fault Diagnosis, Position Mooring 1. INTRODUCTION Ships dedicated floating production and offloading (FPSO) operations in oil exploration and production use surface vessels moored to the seabed to maintain their position. The main objective of such position mooring (PM) is to keep the vessel in a certain position while the secondary objective is to prevent line breakage. A yet unsolved issue is diagnosis of the elements of the mooring system that are submerged several hundred of metres and very difficult to instrument. A deep-water moored vessel has several mooring lines, each consists of the cables, attachment to the vessel, sea bed anchoring and one or more buoyancy el- ements along the line to give the system desired static and dynamic properties (Triantafyllou, 1994). Timely diagno- sis of the state of underwater buoyancy elements would be essential to enhance safe and reliable operation. Regulations normally define different system levels by hardware redundancies, which prevent the system failure by replacing the faulty hardware (DNV, 2011). However the human interactions are significant in fault handling and, hence, the uncertainty introduced in the process of fault recovery is high. Accident databases shows that the human error is the dominant factor when faults develop to failure in maritime accidents (Baker and McCafferty, 2005). Reducing operational risks, and also losses due to unavailability of production systems, could be obtained with timely diagnosis and automated fault-handling of mooring system defects. If diagnosis could be obtained for mooring line buoyancy elements (MLBE), compen- sation actions could be advised or, for systems that also use thruster assistance (Strand et al., 1998; Nguyen and Sørensen, 2009), automatic compensation could be achieved using a set-point shifting algorithm to accom- modate faults in thruster-assisted PM (Fang and Blanke, 2011) and fault-tolerance could be obtained. Systematic fault tolerant control for station keeping of marine vessels was demonstrated in Blanke (2005) where thruster failures were in focus. Structural analysis tech- niques were applied to generate residuals for the fault diagnosis, and control re-configuration design was applied. Sensor failures were dealt with in Blanke (2006) where a full scale validation was also presented. For position moored systems, tension compensation feed-forward con- trol in case of line breakage was suggested in Strand et al. (1998) and line breakage was also the subject in Nguyen et al. (2007), who used switching between pre- designed controllers as fault accommodation. Detection of line breakage was considered by Fang and Blanke (2011) where theoretical considerations by Wang and Xu (2008) indicated that detection schemes should assume a Rayleigh distribution for residuals related to mooring line tension. 8th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes (SAFEPROCESS) August 29-31, 2012. Mexico City, Mexico 978-3-902823-09-0/12/$20.00 © 2012 IFAC 462 10.3182/20120829-3-MX-2028.00224