Uncertainty in Thermal Analysis of Gas Turbine Blades

Pradeep Mohan, Prasanth B. Nair, Andy J. Keane50th 2009 · DOI

Abstract

This paper presents a methodology for thermal analysis of gas turbine blades in the presence of material and geometrical uncertainties. We outline the steps, involving random fleld and spatial discretization procedures followed by the solution and post-processing schemes employing stochastic reduced basis methods, for both types of uncertainties. Numerical studies are presented for a model problem which clearly indicate the speed and accuracy of the method in comparison to benchmark Monte-Carlo simulations. I. Introduction Gas turbine blades in modern aero engines are subjected to tremendous temperatures and pressures. In order to counter the thermal stresses developed due to the high temperature regime, these blades need to be cooled constantly. Hence they are provided with artiflcial cooling through a series of cooling holes by growing the blades as single crystals over a sacriflcial core material which is removed after growth completion, leaving a series of hollow cooling cores in place. Relatively cold air is circulated through these cooling holes resulting in the cooling of the blades. Owing to various factors in the manufacturing process and their operational conditions, the cooling holes vary from blade-to-blade and batch-to-batch due to which the turbine blade geometry varies. Material uncertainties may also be inherent during the crystal growth process. These variations may adversely efiect the life of the turbine blade necessitating maintenance and replacement. Driven by the need to reduce maintenance and warranty costs, improve product quality and reliability, manufacturers are increasingly inclined to innovate optimal designs which are tolerant to geometrical and material uncertainties. Optimization of the blades 1 is only possible when the optimization goals and constraints are quantifled. Quantifying the goals under design constraints in the presence of uncertainties impels probabilistic analysis of the turbine blades. Employing a probabilistic approach, the governing equations with uncertainties or the Stochastic Partial Difierential Equations (SPDEs) are discretized and subsequently solved to obtain the response statistics. Earlier attempts on probabilistic analysis of turbine blades employed Monte-Carlo simulations and response surface methods. 2{4 Monte-Carlo simulations can be computationally prohibitive for large scale and high fldelity models with many degrees of freedom. Response surface methods are a class of approximation techniques that are usually applied to construct linear or quadratic models of the response quantities as a function of the basic random variables. By leveraging the existing literature on function approximations, 5,6 more general nonlinear models can also be constructed. However these methods do not scale very well to problems with a large number of random variables due to the curse of dimensionality which arises from the fact that the number of hypercubes required to flll out a compact M-dimensional space grows exponentially with M. Thus in order to dramatically reduce the computational time for probabilistic thermal analysis of turbine blades, we propose a numerical scheme based on stochastic reduced order projection schemes called Stochastic Reduced Basis Methods (SRBMs), which are signiflcantly faster 7 than other simulation techniques. Firstly, we deal with material uncertainty while conducting thermal analysis of the blades where the thermal conductivity of the blade material is considered to be a random fleld. This random fleld is discretized › u > > ; :
BibTeX
@inproceedings{mohan2009uncertainty,
  author = {Pradeep Mohan and Prasanth B. Nair and Andy J. Keane},
  title = {{Uncertainty in Thermal Analysis of Gas Turbine Blades}},
  booktitle = {50th},
  year = {2009},
  doi = {10.2514/6.2009-2262},
}

Loading related content...

Figure