Ship Performance Monitoring using In-service Measurements and Big Data Analysis Methods
The focus of the project would be to convert the highly dimensional in-service measurement data recorded onboard a ship into meaningful information. Initially, data integrity and quality assurance procedures must be implemented and applied on the data set(s). The cleaned data will be further used to quantify the hydrodynamic performance of the ship. The project will apply big data analysis, data science, and machine learning for data processing. The in-service data will be linked with environmental data from open sources, like for instance NorKyst800 and ECMWF to improve the ability to detect the environmental condition.
Topics like prediction of ship speed loss, speed-powering variation, hull-propeller performance, seakeeping performance, influence of environmental factors will be covered under the hydrodynamics aspect of the project.
- Monitor and optimize the hydrodynamic performance of a ship using in-service measurements from onboard installed sensors.
- Estimation of frictional resistance of the ship from total engine power consumption and ship motions using in-service measurements.
- Estimation of added wave resistance using weather hind-cast data, ship motions and total engine power consumption during voyage
- How to quantitatively represent the hydrodynamic performance of a ship using in-service measurement data?
- How to separate calm water resistance from the total resistance?
- How to convert large amount of sensor data into small number of meaningful hydrodynamic performance parameters?
- How to identify the frictional resistance from the measured total engine power consumption?
- How to estimate sea state and speed-through-water in real time?
Main achievements 2021
- Developed data-driven ship performance monitoring system using the in-service data recorded onboard the ships. Here, calibrated machine-learning models are used to predict the change in performance of a ship through propeller and hull cleaning events.
- Linear machine-learning methods, namely, PCR and PLSR, are shown to be producing comparable results with a well-established non-linear method, ANN, while modeling the hydrodynamic state of a ship using some simple non-linear transformations.
- A semi-automatic data processing framework is developed to process the data from a ship in-service for ship performance monitoring. Several common problems found in the data obtained from a ship in-service are discussed and solutions are suggested.