When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.
Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.
High strength-to-density ratio, high corrosion resistance and superior biocompatibility are the main advantages of Ti-6Al-4V (Ti64), making it a long been favored titanium alloy for aerospace and biomedical applications. Designing titanium components to last longer and refurbishing of aged ones using surface treatments have become a desirable endeavor considering high environmental damage, difficulty in casting, scarcity and high cost associated with this metal. Among mechanical surface treatments, Deep Cold Rolling (DCR) has been shown to be a very promising process to improve fatigue life by introducing a deep compressive residual stress and work-hardening in the surface layer of components. This process has shown to be superior compared with other surface treatment methods as it yields a better surface quality and induces a deeper residual stress profile which can effectively be controlled through the process parameters (i.e. ball diameter, rolling pressure and feed). However, residual stresses induced through this process at room temperature are generally relaxed upon exposure of the components to elevated operating temperatures.
In this work, high fidelity Finite Element (FE) models have been developed to simulate the DCR process in order to predict the induced residual stresses at room temperature and their subsequent relaxation following exposure to temperature increase. Accuracy of the developed models has been validated using experimental measurements available in the literature. A design optimization strategy has also been proposed to identify the optimal process parameters to maximize the induced beneficial compressive residual stress on and under the surface layer and thus prolong the fatigue life. Conducting optimization directly on the developed high-fidelity FE model is not practical due to high computational cost associated with nonlinear dynamic models. Moreover, responses from the FE models are typically noisy and thus cannot be utilized in gradient based optimization algorithms. In this research study, well-established machine learning principles are employed to develop and validate surrogate analytical models based on the response variables obtained from FE simulations. The developed analytical functions are smooth and can efficiently approximate the residual stress profiles with respect to the process parameters. Moreover the developed surrogate models can be effectively an efficiently utilized as explicit functions for the optimization process.
Using the developed surrogate models, conventional (one-sided) DCR process is optimized for a thin Ti64 plate considering the material fatigue properties, operating temperature and external load. It is shown that the DCR process can lead to a tensile balancing residual stress on the untreated side of the component which can have a detrimental effect on the fatigue life. Additionally, application of conventional DCR on thin geometries such as compressor blades can cause manufacturing defects due to unilateral application of the rolling force and can also lead to thermal distortion of the part due to asymmetric profile of the induced residual stresses.
Double-sided deep rolling has been shown as a viable alternative to address those issues since both sides of the component are treated simultaneously. The process induces a symmetric residual stress which can be further optimized to achieve a compressive residual stress on both sides of the component. For this case, a design optimization problem is formulated to improve fatigue life in high stress locations on a generic compressor blade.
All the optimization problems are formulated for multi-objective functions to achieve most optimal residual stress profiles both at room temperature as well as elevated temperature of 450℃. A hybrid optimization algorithm based on combination of sequential quadratic programming (SQP) technique with stochastic based genetic algorithm (GA) has been developed to accurately catch the global optimum solutions. It has been shown that the optimal solution depends on the stress distribution in the component due to the external load as well as the operating temperature.