Ti-6Al-4V (Ti64) is a titanium alloy with a high strength-to-weight ratio and excellent corrosion resistance making it an excellent material for gas turbine applications. The main usage of Ti64 is in compressor blades and discs which their lives are mainly limited by fatigue. Therefore, improving the fatigue life to enhance durability and reliability of such highly stressed expensive component while maintaining the original design can play an important role in design stage and life cycle cost of the engine.
Deep Cold Rolling (DCR) was presented as a promising mechanical surface treatment to improve the fatigue life of components by introducing deep and high compressive residual stresses on the surface and sub-surface layers. However, the exposure to elevated temperatures can relaxes most of the residual stressed induced by the process at the room temperature. Therefore, in order to account for the impact of beneficial residual stress on the fatigue life of the treated component, developing a design tool which can predict the residual stresses induced by the process and after thermal relaxation at elevated temperature is required.
In the present research work, first a high-fidelity FE model was developed to simulate DCR process on a Ti64 plate and the following thermal exposure to elevated temperature 450 ℃. The capability of the developed FE model to predict residual stresses was validated to experimental measurements available in the literature. It has been shown that the finite element predictions correlate well with experimental results with error generally less than 10%. The developed model can be effectively utilized for parametric studies to understand the effect of different process parameters on the
The application of the developed FE model was extended to predict the residual stress profiles in a deep-rolled Ti64 thin plate at room temperature as well as elevated temperature 450 ℃. It was discussed that rolling speed does not have a significant impact on the residual stresses and the number of the rolling pass cannot be considered as design variable. Three key design variables ball diameter (6-12 mm), feed (0.05-0.200 mm) and rolling pressure (10-30 MPa) were considered to form the design domain. DoE was used to discretize the design to main at the design points. The FE models were then run at the design points to generate a data set to train surrogate models using well-stablished machine learning techniques. The surrogate models are considerably lower in order than a full-scale finite element simulation and can efficiently approximate FE models to perform sensitivity analysis and design optimization of process parameters.
Based on the comprehensive literature review, the tensile balancing residual stress created in component was identified as a detrimental secondary effect of the process which results in crack nucleation under HCF regime and needs a particular attention in the design optimization of the process. Operating temperature of the treated component and fatigue behaviour of Ti64 were considered in formulating an optimization problem to achieve the best residual stress profiles in a thin plate using conventional deep rolling. The results revealed that the optimal design variables achieved at the room temperature will not guarantee an optimal solution at the elevated temperature and the operating temperature needs to be considered to derive an optimal solution.
The FE results presented in Chapter 4 show that conventional DCR of thin plate (performing rolling only on one side of the component) can result in unfavorable tensile residual stresses on the untreated side of the components. It was further discussed that the process inherits manufacturing challenges as in can bend and damage thin component. In addition, high gradient asymmetric residuals stress and strain hardening can lead to thermal distortion of the component at elevated temperatures. Double-sided deep rolling was presented as an alternative solution as it treats the both sides of the thin-walled components simultaneously and can be efficiently employed to introduce compressive residual stress on both sides.
A high-fidelity non-linear finite element model has also been developed to simulate the double- sided DCR process on thin Ti64 plate and to predict the residual stress profile introduced by the process and after thermal relaxation due to subsequent exposure to high temperature. The accuracy
of the developed finite element model was validated by comparison with the experimental measurement available in the literature.
The residual stress profile induced by double-sided DCR process is significantly different than that induced by the conventional DCR process performed on the same workpiece under identical processing parameters. The process generates residual stress profiles which are anisotropic and significantly different in axial and tangential directions. Therefore, the stress distribution in the component due to external load should be known prior to selection of rolling direction.
The sensitivity analysis showed that higher rolling pressure and larger ball diameters with lower feed is required to achieve a more compressive residual stress in the axial direction. The tangential residual stress is mainly affected by feed as it significantly influences the tangential plastic deformation and higher feed value are required to achieve more compressive residual stress in tangential direction.
At any given ball diameter, the level of thermal relaxation is higher for deeper and more compressive residual stress which is due to high plastic deformation and thermal instability of dislocations. While the rate of thermal relaxation of axial residual stress is more at lower feed and higher pressure, the tangential residual stress relaxes more when the process is done under higher levels of pressure and feed.
Well-established machine learning principles were then carried out on data generated by the high- fidelity FE model to develop predictive analytical models to approximate residual stresses induced by the double-sided DCR process. The developed analytical functions efficiently replace FE models to perform design optimization of process parameters. Load distribution at high stress areas of a generic compressor blade was considered to formulate a design optimization problem of double-sided DCR process in order to achieve optimal residual stress distributions at room temperature and after thermal relaxation at elevated temperature of 450℃. The optimization study revealed that the operating temperature of the components and the stress imposed by external loading need to be considered in the optimization problem.