Fatigue life of fibre-metal laminates in marine environment

• Main Author: Sai, Wei
• Other Authors: Chai Gin Boay
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Mechanical engineering::Alternative, renewable energy sources; Engineering::Materials::Composite materials
• Online Access: https://hdl.handle.net/10356/145481
• Institution: Nanyang Technological University

This work aims to provide a novel systematic One-stop fatigue model that could predict the fatigue life of the wind turbine blade with fiber-metal laminate (FML) as the shell skin material. Specifically, the fatigue life of FML is linked with a wind blade's lifespan with wind conditions and wind blade airfoil parameters as the inputs and fatigue cycle as the output. Firstly, the wind blade loadings were calculated based on the blade element momentum theory. The aerodynamic loads were calculated and taken as the inputs for airfoil stress analysis, which was further used for stress analysis based on the classical laminate theory. Secondly, the fatigue life was then predicted based on the fracture mechanics approach. Different types of FML, either standard grades or non-standard layups, were fabricated and tested and involved in the fatigue prediction model. The predicted S-N curves were validated with the experimental data. The comparison shows a good correlation, and the proposed One-stop model has achieved a reliable accuracy within one order of magnitude. Further, to enhance the classical fatigue life prediction in generalization, an intermediate gray box model between the expert system and the black box system was proposed, a regression tree ensemble-based (RTE) machine learning approach. Unlike the total black box system, this grey box model provides an interpretable and extendable solution towards various FML layups. What’s more, the RTE model overcomes the limitation of classical fatigue theories, i.e., lack of generalization and extensibility. Specifically, in the gray box RTE model, mechanical, geometrical properties, and fatigue loading stresses are selected as the training parameters (so-called features). FML fatigue life is predicted as the output of the model. Experimental data of 128 pieces of FML specimens with different layups were used to train and validate the machine learning model. Results show that the model can provide better fatigue life prediction accuracy and model reliability than the classical fatigue model. The most correlated, either positively or negatively, parameters to the fatigue life span are the stress developed in the aluminum layer, the maximum cyclic stress, alternating stress, and mean fatigue stress. To exclude the influence of human errors in the whole experiment process, the RTE model was further trained based on the analyzed data obtained from the One-stop model. The model accuracy and reliability in terms of mean absolute error and the determinant of coefficient, respectively, have shown significant improvement.