Bayesian Framework for Gas Turbine Simulator with Model Structural Uncertainty
Piyush Tagade, Kumbhar Shrikant Sudhakar · 50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference · 2009
The paper discusses a Bayesian framework for calibration of a gas tur-bine simulator in presence of uncertainty in the model structure. The framework is proposed for inference of a compressor map. Combined uncertainty in the compressor map and model structure is probabilisti-cally specified using Gaussian stochastic process. Markov Chain Monte Carlo (MCMC) method is used to sample from the posterior distri-bution. The proposed framework is demonstrated for simulation of a single spool turbojet engine with artificially introduced uncertainty in the model structure. Nomenclature A = area, m2 CN = corrected speed f = probability density function F = thrust, N H = altitude, m I = polar moment of inertia, kg.m2 M = flight Mach number N = spool speed, rev/s P = pressure, Pa PR = pressure ratio Pr = probability R = gas constant, J/kg.K T = simulator TR = temperature ratio V = lumped volume, m3 W = mass flow rate, kg/s x = vector of control input Y = system response α, λ = hyperparameters of the Gaussian process γ = isentropic index ǫ = measurement uncertainty