Highly-loaded low-speed roller bearings are commonly used in the offshore machinery, for example in turret moorings systems of Floating Production Storage and Offloading (FPSO) units and in slew bearings in heavy-lifting cranes. Safe and reliable operation and structural integrity can be ensured by monitoring the condition of the bearing. Condition monitoring based on acoustic emission (AE) has shown good potential for monitoring the health of highly-loaded low-speed bearings. With the increase in potential and proven condition monitoring methods, the need for lifetime predicting methods is increasing.
In this thesis, the feasibility of a digital twin (DT) concept is tested to make lifetime predictions of a highly-loaded low-speed roller bearing. The developed DT uses a combination of the Paris model and an AE-based model presenting the current state of the bearing. In the remaining useful life (RUL) prediction, the final crack size and total number of load cycles are determined based on an integrated form of the Paris model. Three different forms were tested for the RUL prediction.
The Paris model and the AE-based model showed good agreement in a benchmark case, based on which the two models were combined. The AE-based prediction was tested for hits and counts, with the counts yielding into better results due to a better fitting of the model constants. The linear RUL based on the number of load cycles resulted in the best prediction. The developed DT showed good potential in making lifetime predictions of highly-loaded low-speed roller bearings and enrichment of future DTs with prognostic information.