Hedengren, J.D., Brower, D.V., Kidder, K., Hillman, Z., Data-Driven TLP Tendon Loads from Internal Hull Fiber-Optic Sensors, ASME 42nd International Conference on Ocean, Offshore and Arctic Engineering, OMAE2023/103309, Melbourne, Australia, June 2023.
Preprint: apm.byu.edu/prism/uploads/Mem...
Presented at OMAE 2023
A new and innovative monitoring system is based on fiber-optic sensors and has been developed and installed on a Tension Leg Platform (TLP) in the Gulf of Mexico. The system design is implemented to replace standard load cells that have been the workhorse of providing tendon load data over the last two decades. Many load cells have been efficient but have reached the end of life and on most TLP’s have either failed or have degraded performance.
The new monitoring system is based on a data-driven models to infer tendon load data from internal hull fiber-optic sensors. There are no subsea parts and all sensors, cabling, and hardware is located topside. This method involves several hundred Fiber Bragg Grating (FBG) sensors that are placed on a webframe within the hull of the TLP. The location of the sensors is based on a detailed Finite Element Analysis (FEA) and maximized correlation between webframe strain and tendon load. One year of fiber-optic sensors data (700 GB) trains regression models that correlate webframe strain to tendon loads. The new sensors and predictive models provide measurements within 0.8% mean absolute percentage error (MAPE) of the original load cells over the course of several months. This equates to +/- 40 kips (40,000 lbf) on each of the six tendons during wave action, tide fluctuations, storms, and strong loop currents. A resolution of 100 kips (100,000 lbf) is required for verification within a weight triangle, showing the feasibility of this system to replace the load cells.
There are several new innovations and features with this new system that are reported here. The sensing system is dry-installed in the ballast tanks of the TLP with sensor orientation designed to maximize correlation between tendon load and fiber-optic sensor strain variation. All electronic components are topside with cabling penetrations into the ballast tanks. The fiber-optic sensors are multiplexed together to minimize cabling and maximize sensor count. These fiber-optic sensors immune to electromagnetic interference and corrosion. The data-driven model from the fiber optic sensor data is integrated directly into the existing IMMS (Integrated Marine Monitoring System). There is direct correlation of tendon loads with the new sensors when trained with the remaining load cell sensors. The fiber-optic sensors collect strain measurements at 1000 Hz and report new load measurements with 100 measurements consolidated into one load value sent to the IMMS every 0.1 seconds. There are consistent data measurements and reliable analysis of peak displacement from each of the sensors with resolution less than 0.1 nm. For comparison, the diameter of an iron atom is 0.126 nm. The data is divided into training and test sets with one year of training and months of validation. The training and validation includes assessment of environmental effects such as wave action, tide cycles, storms and excessive weather, and annual water temperature changes in the Gulf of Mexico. Fiber-optic temperature gauges are installed on the webframes and compensation is built into the regression algorithm.