Robust Design Optimization for Fatigue Life with Geometric and Material Uncertainties of Mechanical Parts Under Random Loading Based on Maximizing Fatigue Life and Minimizing Uncertainty in Fatigue Llife Prediction

Esfahani, Saeed | 2019

604 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 52351 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Khodaygan, Saeed
  7. Abstract:
  8. Fatigue life prediction of a mechanical part is one of issues which a group of engineers are engaged with it and always they try to design the parts with the maximum of lifetime. Although many researches have been done in this field but yet we can see that predicted life are different from that happens in the reality because there are some uncertainties in the phenomena. Our effort in this project is creating an algorithm design so that the parts are designed by it, have the maximum fatigue life and the minimum uncertainty in prediction. In this project we have considered geometrical, material and random loading uncertainties as error resources. Older methods those are presented in this field have some defects and we have tried to solve them by a new algorithm. In this algorithm first we should specify design and random variables of the problem and then according to our fatigue problem we must choose an appropriate fatigue life prediction method. Then if we are not able to calculate the life by an explicit relation that in it, life is a function of design and random variables, we will calculate the life by finite element method in several points and afterwards we will use a Metamodel method to create an explicit relation between fatigue life and design and random variables. We must find those life calculating points by a suitable sampling method. With this relation we can utilize dimension reduction method to calculate mean and standard deviation of the life at any arbitrary point. Then we define an optimization problem to maximize the men and minimize the standard deviation. In this stage we can add some arbitrary goals to the design problem. By solving the optimization problem using multi objective genetic algorithm we will find several non-dominated design points (Pareto Frontier). We use TOPSIS method to pick the best point for design variable.Three design problem examples have been solved to check efficiency of the introduced algorithm. In these problems, we put different challenges like calculating the life analytically or by using finite element method, existence of several types of distribution and so on. It has been shown that this algorithm in aspect of computational cost is so better than other methods and its precision in statistical moments estimation is good
  9. Keywords:
  10. Fatigue Life Reduction ; Robust Optimization ; Dimensionality Reduction ; Metamodel ; Genetic Algorithm ; Random Loading

 Digital Object List