Study of the dependencies between in-service degradation and key design parameters with uncertainty for mechanical components

The design features of machine components can impact significantly in its life while in-service, and only relatively few studies which are case specific have been undertaken with respect to this. Hence, the need for more understanding of the influence of geometric design features on the service life of a machine component. The aim of this research is to develop a methodology to assess the degradation life of a mechanical component due to geometric design influence in the presence of uncertainties and its application for the optimisation of the component in the presence of these uncertainties. This thesis has proposed a novel methodology for assessing the thermal fatigue life, a degradation mechanism based on the influence of design features in the presence of uncertainties. In this research a novel uncertainty analysis methodology that is able to handle simultaneously the presence of aleatory and epistemic uncertainties is proposed for a more realistic prediction and assessment of a components thermal fatigue degradation life estimated using finite element analysis. A design optimisation method for optimising the components design in the presence of mixed uncertainty, aleatory and epistemic uncertainties is also proposed and developed. The performance of the proposed methodology is analysed through the use of passenger vehicle brake discs. The novel uncertainty quantification methodology was initially applied on a solid brake disc, and validated for generalisability using a vented brake disc which has more complex design features. While the optimisation method as proposed was applied on the vented brake disc. With these this research proposes a validated set of uncertainty and optimisation methodology in the presence of mixed uncertainties for a design problem. The methodologies proposed in this research provide design engineers with a methodology to design components that are robust by giving the design with the least uncertainty in its output as result of design parameters inherent variability while simultaneously providing the design with the least uncertainty in estimation of its life as a result of the use of surrogate models.