Effect of Input Variability on the Performance of Turbine Blade Thermal Design Using Monte Carlo Simulation: An Exploratory Study

Abstract

A probabilistic approach to the thermal design and analysis of cooled turbine blades is presented. Various factors that affect the probabilistic performance of the blade thermal design are grouped into categories and a select number of factors known to be significant, for which the variability could be assessed are modeled as random variables. The variability data for these random variables were generated from separate Monte Carlo simulations (MCS) of the combustor and the upstream stator and secondary air system. The oxidation life of the blade is used as a measure to evaluate the thermal design as well as to evaluate validity of the methods. Two approaches have been explored to simulate blade row life variability and compare it with the field data. Field data from several engine removals are used for investigating the approach. Additionally a response surface approximation technique has been explored to expedite the simulation process. The results indicate that the conventional approach of a worst-case analysis is overly conservative and analysis based on nominal values could be very optimistic. The potential of a probabilistic approach in predicting the actual variability of the blade row life is clearly evident in the results. However, the results show that, in order to predict the blade row life variability adequately, it is important to model the operating condition variability. The probabilistic techniques such as MCS could become very practical when approximation techniques such as response surface modeling are used to represent the analytical model.