Exploring the use of Adjoint Methods for Detailed Sensitivity Analysis on Turbomachinery

CATEGORY: TURBOMACHINERY
EngOpt 2012 - International Conference on Engineering Optimization, Rio de Janeiro, Brazil, 1-5 July 2012.
Exploring the use of Adjoint Methods for Detailed Sensitivity Analysis on Turbomachinery
Andre C. Marta and Sriram Shankarany
andre.marta@ist.utl.pt
shankaran@ge.com
Instituto Superior Tecnico, Lisboa, Portugal
Global Research Center, General Electric, Niskayuna, USA
Abstract
Significant progress has been made in the  field of optimization using high- fidelity models. The increased use of adjoint methods has allowed the computation of sensitivity information, required by gradient-based optimizers, in a very efficient manner. In terms of fluid dynamics, while the  first applications were focused on aerodynamic shape design, recent approaches to the development of adjoint solvers, namely with the use of automatic differentiation tools, have made it possible to extend their capabilities far beyond that.
Authors describe a discrete adjoint method implementation for a generic CFD solver and lays down the steps to estimate sensitivities of functions of interest with respect to any variables handled by the flow solver. The applications presented are based on turbomachinery blade design problems. Two different capabilities are illustrated: one more traditional geared toward shape optimization that focuses on estimating gradients of some turbomachinery aerodynamic performance parameters with respect to blade geometry, and another more innovative geared toward estimating the effect of the inlet or exit boundary conditions on some aerothermal performance parameters. The computational cost required by this method in terms of CPU time is considerably reduced compared to the popular  nite-difference method, at the expense of a more complex implementation. The detailed sensitivity information obtained is discussed from a designer perspective. Other possible applications of adjoint methods are listed and their development implications are also described.
Free Full Text Source: http://www.engopt.org/authors/title.html

Inverse Computation Scheme of Turbomachinery Blade Shapes Applied to Axial Hydro-Turbine Runners

CATEGORY: TURBOMACHINERY
EngOpt 2012 – 3rd International Conference on Engineering Optimization, Rio de Janeiro, Brazil, 01 - 05 July 2012.
Inverse Computation Scheme of Turbomachinery Blade Shapes Applied to Axial Hydro-Turbine Runners
Marcos Antonio Rodrigues dos Santos, Nelson Manzanares Filho, Waldir de Oliveira, Luis Guilherme Cunha Santos
ars.marcos@unifei.edu.br
nelson@unifei.edu.br
waldir@unifei.edu.br
lgcunhas@gmail.com
UNIFEI – Federal University of Itajubá, Itajubá-MG, Brazil
Abstract
Proposes an inverse scheme for computing the mean through flow surface of turbomachinery. The flow is considered inviscid and incompressible with the blades supposed infinitely thin. The computational domain consists of a meridional projection extending from the upstream section to the downstream section. Averaging the continuity and momentum equations leads to an equivalent elliptical differential equation for the streamfunction
which is complemented by (i) the conservation of hydraulic torque condition along streamlines for the bladeless regions and (ii) the mean flow tangency condition for the bade region. The blade region may be treated in direct mode solving for the streamfunction and hydraulic torque with a prescribed blade shape. For design purposes one adopts here an inverse approach by solving for the streamfunction and blade shape with a given hydraulic torque distribution. The streamfunction equation is solved iteratively by a differential quadrature method with Successive Over-Relaxation (SOR) using a non-orthogonal mesh. The projections of the blade leading and trailing edges coincide with mesh knots. An application is made for designing the conceptual mean runner blade surface of an axial flow hydro-turbine.
Introduction
With the progress of Computational Fluid Dynamics (CFD), the analysis of complex flow that develops inside the axial hydroturbine can be solved by using the Navier-Stokes or the Euler equations. Despite the increasing use of CFD techniques for the design and analysis of individual components or complete turbines, several flow types related with the boundary layer separation, vortex formation, interaction between the flows, cavitation, among others, in general can arise even at the point of the project.
Free Full Text Source: http://www.engopt.org/authors/title.html

Risk assessment of failure modes of gas diffuser liner of V94.2 siemens gas turbine by FMEA method

CATEGORY: TURBOMACHINERY
25th International Congress on Condition Monitoring and Diagnostic Engineering, Journal of Physics: Conference Series Volume 364, 2012 J. Phys.: Conf. Ser. 364
Risk assessment of failure modes of gas diffuser liner of V94.2 siemens gas turbine by FMEA method
H Mirzaei Rafsanjani and A Rezaei Nasab
mirzaei.hesam@mapnaturbine.com
rezaei.alireza@mapnaturbine.com
Mapna Turbine Engineering & Manufacturing Company (TUGA), Mapna Building, No. 231, Mirdamad Ave, Tehran 191895365, Iran- P. O. Box: 15875-5643
Abstract
Failure of welding connection of gas diffuser liner and exhaust casing is one of the failure modes of V94.2 gas turbines which has occurred in some power plants. This defect is one of the concerns of customers which must be addressed before they are willing to accept the final commissioning of this product. Accordingly, the risk priority of this failure was evaluated by failure modes and effect analysis (FMEA) method to find out whether this failure is catastrophic for turbine performance and is harmful to humans.
By using history of 110 gas turbines of this model which are used in some power plants, the severity number, occurrence number and detection number of failure determined and consequently the Risk Priority Number (RPN) of failure determined. Finally, critically matrix of potential failures is created and illustrated that failure modes are located in safe zone
Free Full Text Source: http://iopscience.iop.org/1742-6596/364/1/012137