Condition monitoring of fibre ropes using machine learning

Abstract

The application of fibre ropes in offshore lifting operations has significant potential for further development. With minimum breaking loads (MBL) equivalent to steel wire at similar diameters and almost neutral buoyancy in water, it is in theory possible to reach depths exceeding 3000 m with smaller cranes and vessels, representing substantial savings in not only potential operation costs. However, with fibre ropes there are different requirements and standards to consider with regards to condition monitoring, maintenance and retirement criteria. Safe and reliable operations are paramount in the offshore sector and any incidents that occur during offshore lifting would not be only significantly damaging financially but could potentially lead to loss of life. Current standards for fibre rope condition monitoring originate in mooring applications, and are based on manual inspection for retirement and re-certification. There is significant room for developments in methods that can aid the inspection process. To address this problem, computer vision and thermal monitoring methods for fibre ropes are developed and experimentally investigated at the Mechatronics Innovation Lab in Grimstad, Norway. The methods are used to monitor changes in fibre rope condition during cyclic-bend-over-sheave testing and to find relevant condition indicators that give more information regarding the condition and remaining useful life of the fibre rope. In addition, the data recorded is used to form machine learning models that both classify rope condition and predict the remaining life of fibre ropes during CBOS testing. The expected outcome is to use physics-based machine learning methods to improve both condition classification and remaining useful life estimation of bre ropes used in offshore lifting operations. In the appended papers at the end of this thesis, the proposed methods have been experimentally investigated and validated through cyclic-bend-over-sheave experiments performed at the Mechatronics Innovation Lab and further data analysis performed at the University of Agder, Norway and at divis in Dortmund, Germany.