Artificial Intelligence and Mathematical Modelling for Riser Integrity Structural Evaluation

Artificial intelligence (AI) offers large opportunities for efficient drilling and completion of oil and gas wells as well as for subsequent production. Sensors on offshore vessels produce a large amount of data that can be used to optimize systems performance during all phases of offshore operations. Reducing cost and increasing uptime are key targets, but monitoring systems may also contribute to increasing safety and reducing risk of serious incidents offshore. Artificial Intelligence in combination with mathematical modeling may assist in monitoring processes and advising on appropriate actions at the right time. Fatigue of riser and wellhead systems during drilling and completion has been a challenging problem for operators. Larger blow out preventers (BOP) have been introduced to enhance safety on drilling rigs, but an undesired side effect has been larger loads on the wellheads and increased fatigue. Monitoring systems have recently been widely applied to monitor loads on the wellheads. Mitigating measures have been applied to relieve the wellheads of the loads. The Reactive Flex-Joint (RFJ) is a TechnipFMC product that has proven effective and easy to use. The study focuses on the use of mathematical modelling of riser mechanics using different sensors readings to estimate loads on the wellhead and the lower section of the riser. Machine learning in combination with conventional models is used for estimating loads on wellheads and risers.

Drilling risers are slender structures used during offshore operations by drilling companies to provide access from a surface vessel to the top section of the well (wellhead) at the sea floor. The purpose is to allow for drilling or completion of wells or to perform workover of the wells at a later stage. The lowermost section of the riser is usually a blowout preventer (BOP) that is to contain hydrocarbons in the case of an unexpected release. However, BOPs are very heavy structures that cause significant structural fatigue of wellheads when exposed to loads from the riser. Different technologies have been used to reduce the loads applied to the wellheads. The study focuses on a few much used technologies including the Reactive Flex-Joint (RFJ) which has been in commercial use for approximately half a year. The RFJ is effectively a muscle placed on top of the BOP counteracting the loads applied by the riser. The technology has been carefully studied earlier, but new data allows for more in-depth understanding of its complex mechanical response. Global riser analysis models (numerical models) are much used to assess structural integrity of wellhead and risers. There are in the literature, however, few examples of such models being independently validated by measurements. Thus far it has not been common to place sensors on riser systems to measure loads directly due to relatively high cost of sensor systems. Furthermore, systematic evaluation of data from sensors is not widely reported. Most models used for assessment of riser loads are based on finite element methods or other types of discretization of the fundamental partial differential equations. An alternative approach however are machine learning algorithms that rather make use of measured data from operations offshore. Such methods could in principle be more accurate and faster when used within certain limits.