Civil aviation traffic will increase in the future, so reduce:

• direct operating costs

• fuel burn

• emissions

• noise

Need to improve design technology in:

• Motor

• Propulsor

Improve motor thermal efficiency by:

• Increase TET

• Increase OPR

Future architectures:

• Higher BPR (currently 11, will be 15+)

• Lower FPR

This will:

• Lower specific thrust

• Improve propulsive efficiency

Why is the exhaust important?

• Increasing BPR will increase gross to net propulsive thrust

• Designs are therefore more sensitive to variations in gross propulsive thrust

• Gross propulsive thrust is linearly dependent on the aerodynamic performance of the exhaust

Which components will be analysed?

• Bypass duct

• Nozzle

• Post exit components

What is separate jet exhausts?

• Core cowl separates core flow from bypass flow

• Protruding core plug limits the length of core cowl

What is the problem with separate jet exahusts?

• Can be substantial sources of thrust loss (gross thrust reduced by 2%)

How is performance measured?

• Discarge coefficient

• Velocity coefficient

CFD is a reliable performance prediction tool

It is also efficient and used in design optimisation (Efficiency is depending on mesh size, geometry complexity, schemes, solvers and models used)

Aerodynamics of the exhaust is important for future high BPR engines

What is unique about the current work?

• Previous authors have looked at optimising exhaust nozzles

• A holistic approach for separate jet exhausts including bypass, core duct and post exit components has not been reported

• Impact of high BPR engines and lower FPR on exahust system design and optimisation has not been reported

What is the approach?

• Cycle analysis

• Geometry parameterisation

• Mesh generation

• RANS flow solution

What is new?

• Expand optimisation strategy using DOE (Design of Experiment), RSM (Response Surface Modelling) and GA (Genetic Algorithm)

What is being optimised?

• Current and future engine architectures

• Large turbofans

• Optimise the exhaust designs

What is GEMINI?

• Tool developed is GEMINI

• Designs complete exhaust system for designated engine cycle using key engine hardpoints

• Applicable to:

  • Engine performance simulation

  • Exhaust duct and nozzle aeroline parameterisation

  • Viscous compressible flow

What is the process?

• Designate a set of thermodynamic cycles and geometric design parameters

• Analyse engine at design point and off design (0D conditions) - Turbomatch, output bypass and core sizes and flow capacities, at steady state conditions

• Inverse design to create 2D axi-symmetric model

• Automatic generation of grid

• Convege CFD solution

• Determine discharge and velocity coefficients

How is the parametric geometry defined?

• Kulfan’s CST functions

• Qin’s CST variations

• Bypass, core, duct exhaust are reduced to a set of analytical expressions

• The expressions are functions of a standard set of design parameters

How is the nozzle designed?

• Geometric throat area is known

• An effective convegent-divergent ratio is defined

• Application of the rolling ball area estimation method to nozzle exit plane and upstream CP results in a series of control points that satisfy the prescribed design parameters

How is the upstream duct defined?

• Direct control of a series of control points

Why is the engine intake considered?

• To capture the effect of inlet mass flow capture ratio

• To then account for the effect of the static pressure distributionon the nacelle

• To then account for the effect of freestream supression on the aerodynamic performance of the exhaust system

How is the geometry defined?

• Upstream duct via specifying position, slope and curvature within a series of control points

• Core cowl and plug are modelled as straight lines

• Includes a third nozzle

What is done in this paper?

• Extend GEMINI

• Implement DSE and optimisation environment

• Non-linear nature must be dealt with

• Must mitigate the cost of numerous CFD applications

How is the process of DSE done?

• Deployment of DOE method to explore the available design space

• Construct RSMs from DOE results

What kind of DOE is used?

• Latin Hypercube

What is a RSM?

• Hypersurface describing the mathematical relationship between a set of imposed design inputs and outputs

• The use of RSMs will avoid a prohibatively large number of CFD simulations

• Interpolation using Gaussian process regression, Kriging interpolation

• Performance metrics are discharge and velocity coefficient

• Leave-one-out cross validation used to check predictive accuracy of RSMs

How is the optimisation done?

• Global method to avoid being trapped in locally optimal solution - GA (Genetic Algorithm)

How are the baseline engines defined?

• Optimise low pressure exhaust system design and core afterbody aerolines for current and future aero-engines.

• BPR current = 11

• BPR future = 15

• OPR, TET, component efficiencies selected according to technology guidelines

• Each cycle optimised wrt FPR to maximise specific thrust

• 2D axi-symmetric

• Geometry from public domain

• Predictions at mid-cruise

• Bypass is choked, core is unchoked

How are the parameters designed?

• 11 to 12 variables for future and current engines

• Outer line angle is kept constant for future engine

How are the RSMs constructed?

• Using DOE data

• Interpolation using Gaussian processes regression (Kriging interpolation)

• Quadratic regression function and squared exponential autocorrelation function

How are the RSMs checked?

• Leave one out cross validation

• Employs all the avaliable data apart from one, which is the one to prediction

• Prediction is compared with original raw data for accuracy

• Surrogate model predictions are correlated against raw data using Pearson’s product moment of correlation

• Also assesses averge model error and standard deviation for each performance metric

Result shows that CFD raw data and predicted data has very high correlation.

Could be improved using a larger amount of data

Low percentage error

Standard deviation is of similar order to error, so data is scattered - shows non-linearity of the system

How is the optimisation performed?

• Genetic algorithm

• Advantage of using RSMs is that they are more efficient than CFD models

What is the process for the GA?

• For current and future engines

• Optimise in terms of overall velocity coefficient

• Population size is 10 times number of design variables

• 40 generations

• Convergence criterion of \(10^{-12}\)

What are the results of the optimisation?

• Good solution achieved within 500 evaluations

• Still contains small number of unfit individuals

• Improvement wrt baseline values is large (2-4% in thrust)

• CP to exit length ratio is increased (as before) mitigating strong shock

• Also flow separations are mitigated