Title : Fatigue behavior of polychloroprene rubber (CR) with tungsten nano-particles considering stress triaxiality based on semi-empirical and machine learning models
This paper presents a strain-based semi-empirical fatigue life prediction model of tungsten (W) nano-particles reinforced polychloroprene rubber (CR) material (W-CR material) developed for nuclear decommissioning manufactured by rolling process. The developed semi-empirical fatigue life prediction model is based on Mason-Coffin formulation considering the anisotropic mechanical behaviour and the stress triaxiality. To consider various multiaxial stress states, a novel fatigue experiment method which utilizes limiting dome height test (LDH test) has been developed. In addition, three types of specimens have been cut along 0º (rolling), 45º and 90º to consider the W-CR material’s anisotropic behaviour. In the semi-empirical model, the effect of stress triaxiality is characterized as the form of an exponent function. Moreover, the anisotropy is represented by the ratio between the maximum stress of the specimen made in the rolling direction and each direction. The developed model is able to consider low cycle fatigue (LCF) and high cycle fatigue (HCF) together. The semi-empirical fatigue life prediction model shows a high correlation factor equal to 91.4%. Furthermore, a machine learning (ML) models have been applied for the fatigue life prediction. The accuracy of the adopted ML models with five training set compositions have been compared. The machine learning method shows a lower mean relative error than the semi-empirical model. The advantage of the proposed models in this paper is the anisotropic behavior of rubber composites can be considered easily. It is concluded that the proposed models are reliable and useful to predict fatigue life for various stress states of rubber composites.