Loading...

Reliability Based Robust Design Optimization of Mechanical Systems with Uncertain Parameters Using Bayesian Inference

Hassani, Hossein | 2018

1058 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 51302 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Khodaygan, Saeed; Asempour, Ahmad
  7. Abstract:
  8. This world’s commercial and competitive space force’s designers and manufactures to produce and supply products with high quality and low price at desirable level of reliability. On the other hand, during the design and production process, Engineers are always faced with uncertainty. Essentially, uncertainties can be divided into two categories: aleatory and epistemic. Aleatory uncertainties are due to changes in design conditions, material properties, physical dimensions of components and operating conditions. This kind of uncertainty will not be reduced during the life.on the otherhand, Epistemic uncertainties that are due to lack of knowledge about physical nature of systems can be reduced by collecting data. However, collecting data can be time-consuming and costly. To deal with uncertainties, several tools have been suggested in litreture: robust design optimization (RDO), reliability based design optimization (RBDO) and reliability based robust design optimization (RBEDO). The main objective of RDO is to improve quality by minimizig the design sensivity with respect to variations. In RBDO, the goal is to find optimum design while maintaining at desirable level of feasibility. In recent years, RBRDO algorithms have been improved to ensure quality and reliability of products simultaneously. The main challenge of existing RBRDO methods is computational cost, since RBRDO is a multi-objective optimization problem with sophisticated computing. In practical engineering problems, uncertainties of some design or parameter variables are epistemic and information about them are available in the form of limited samples. Generally, Existing RBRDO methods have taken two approachs to deal with these kind of uncertainties; some of them don’t consider these informations in design process. So they lose these valuable informations. Other groups assumes these uncertainties distributions and then estimate their parameters. This approach can lead to an error. In this research, we have proposed a comprehensive RBRDO framework by combining Bayesian reliability analysis and DRM method by employing NSGA2-II multi-objective optimization algorithm. In this framework, the design problem will be converted into a three objective optimization problem with minimizing cost function, minimizing variance of cost function as robustness measure and maximizing reliability of system objectives. To reduce computational costs, in addition to first order reliability method, eigenvector dimension reduction method has also been used in this research. In the proposed framework, To analyse the final pareto fronts, topsis methods is employed and the most approtriate solution is selected.To verify the application of proposed algorithm, three engineering example is solved and their solutions are discussed
  9. Keywords:
  10. First Order Reliability Method (FORM) ; Multiobjective Optimization ; Optimal Design ; Robust Design ; Non-Dominate Sorting Genetic Algorithm (NSGAII) Method ; Bayesian Inference ; Eigen Vector Dimension Reduction (EDRM)

 Digital Object List

 Bookmark

No TOC