During installation of the Big Foot TLP, nine of the 16 pre-installed tendons fell to the seafloor prior to completion of installation. a comprehensive Root Cause Analysis was undertaken to determine the "cause" of the incident. During the early investigative phase, a number of sources of data were identified. Two particular information sets were gathered, interrogated, processed, etc. to extract usable patterns and values to aid in decision making. A third composite data set of deepwater current values was coupled to extract additional insights. All data sets were designed for installation support and not intended for a forensic investigation. Identification of the data, understanding original intent, multi-stage filtering and conclusions will be illustrated.
One is a time series containing x, y, z spatial data of a common location on each of the 16 pre-installed tendon assemblies and the second being non-periodic discrete air volume measurements of buoyancy devices. The time series data contained repetitive data points, contained spurious readings, etc. The multi-stage automatic and manual filtering processes utilized to develop non-periodic traces useful for insight and decision making is described. Accompanying this discussion will be the description of original intent and the adaptation from intended use to RCA value. The second data involved non-periodic air volume measurements. Initial interpretation of data during installation was sufficient for operational purposes. A Root Cause Analysis team analyzed the data including use of correlative time series unrelated to the measurement to identify telling patterns useful in decisions.
The absolute as recorded data was unable to provide sufficient confidence using standard data processing techniques. Target manual filtering and correlative comparisons provided useful and valuable decision-making insights. The ability to use data designed for specific purposes not aligned with ultimate usage for RCA purposes is shown