An Improved Process for Mapping Aeroelastic Loads Across Structural Meshes

Douglas J. Neill, Jack F. Castro, Patricia E. Jones

MSC.Software Corporation

Agenda

• Typical loads processes and issues• Proposed process• The ADB/AEDB collections• Methodology• Results• Summary

An Aeroelastic Loads Process

CFD or WT

“Panel Code”

Rigid Aero Loads

AIC

“Flexibility” FE

Corrected LoadsOn

Coarse FE

Corrected LoadsOn

Internal Loads FE

AeroelasticSolver

LoadsTransfer

Tools

Critical Loads SurveyStress GroupOther tools

Issues with Typical Process

• 2 common approaches to transfer coarse loads to the internal loads model– Smearing component integrated loads

• Heuristic• Prone to error• Slow, engineer-in-the-loop • Loss of balance during smear• Re-balancing is itself heuristic and slow

– “Common Loads Points”• Difficult to coordinate• Tends to degrade FE model quality• Degrades computational performance

Issues with Typical Process

• The outlined approach works, but it’s a problem in many organizations– Too slow

• Cannot keep up with critical loads survey needs

– Too easy to make mistakes– Cannot be repeated reliably

• Too many engineer-in-the-loop heuristics• Too easy to lose the association of loading event (on

coarse FE) to critical load (on internal loads FE).

The Proposed Process

• “Remember” the elastically corrected loads on the aerodynamic mesh– As MSC.FlightLoads does now on the ADB and

AEDB

• Re-evaluate the mapping of these load increments onto the new structural FE– Re-use standard coupling methods

• Gather the loads onto an updated AEDB for use downstream– Downstream tools don’t realize a two-step

process was applied upstream

The Proposed Process

MSC.FlightLoads“AEDB”

(Coarse FE)

CFD or WT

“Panel Code” MSC.FlightLoads“ADB”

“Flexibility” FE

AeroelasticSolver New

MappingTool

“Internal Loads” FE

SplineTo

New FE

MSC.FlightLoads“AEDB”

(Internal Loads FE)

Critical Loads SurveyStress GroupOther tools

Can be applied recursively

The ADB/AEDB Collections

• The ADB contains the aerodynamic model– Mesh topology– Rigid aerodynamic loads

• The AEDB contains the ADB and “flexible increments” (FI)– FI include forces and displacements– Forces arise from rigid and from the

displacements passing thru the AIC– All forces (except rigid inertial) are on both the

aero and structural meshes– These are “unbalanced”, but all balanced

states are a linear combination

Mapping Methodology

• From equations of motion for static aeroelasticity1. rigid load + AIC deformation2. AIC * deformation elastic increment force3. rigid load + elastic force corrected load

• From 2 on the aero mesh, we map to the new structural FE via DMAP alter (for now)– From these loads, a statics solution on new FE

yields deformations on new FE– Restore inertial loads (rigid) from new FE– Proceed with normal AEDB creation to complete

the AEDB collection

Advantages

• No heuristic methods– Two mappings of the same kind– Repeatable– Recursively applicable

• Full vehicle balanced loads are immediately available– Trim on the new structure– Using mass of the new structure

• BUT using inertial aeroelastic correction of the original FE

• FASTER for superior FIDELITY

An Example

The common aero mesh

The built-up structure

The beam-stick structure

Cases• All analyses represent a 1g Level Flight

Trim at M=0.4, Sea Level• 3 variants were run

– Aero + Beam FE– Aero + Built-up FE– Starting with Aero + Beam, Map to Built-up FE

and then trim• Model Sizes

– Aero: 554 Boxes– Beam FE: 141 nodes– Built-up FE 3171 nodes

• All runs were made on NEC laptop, Pentium II

Results and Timing Summary

  Beam FE Solution Built-up FE Solution Beam Solution Mapped to Built-up

ANGLEA1.740 degrees

1.702 degrees 1.740 degrees

SIDES 0.300 degrees 0.338 degrees 0.211 degrees

RUDDER -0.552 degrees -0.623 degrees -0.603 degrees

ELEV 8.311 degrees 8.159 degrees 8.310 degrees

AILERON(s) 0.450 degrees 0.292 degrees 0.463 degrees

CPU (ADB/Mach) 13.1 13.1 0 (taken from ADB)

CPU (Structural) 22.2 32.3 29.1

CPU (AEDB/Mach&Q) 13.4 128.8 16.7

CPU (Total) 49.0 174.4 46.5

Results and Timing Summary

0500

10001500200025003000350040004500

CPU

1 2 3 4

Increasing Mach/Q Pairs

Comparison of CPU Growth

Total Direct

Total Remapped

1/1 2/2 2/6 5/30

Number of Mach/M-Q Pairs

Qualitative Results

Trimmed ForcesOn

Aero Mesh

Trimmed ForcesOn

Beam Mesh

Qualitative Results (cont.)

Trimmed ForcesOn

Aero Mesh

Trimmed ForcesOn

Built-up Mesh

Quantitative Results

Comparison of Bending Moment

-20000

0

20000

40000

60000

80000

100000

120000

0.00 50.00 100.00 150.00 200.00

Span Station

Ben

din

g M

om

ent

Aero Mesh BendingMoment

Beam Bending Moment

Refined FE BendingMoment

Zoom into Tip Area

Quantitative Results

Comparison of Tip Bending Moment

-500

0

500

1000

1500

2000

150.00 155.00 160.00 165.00 170.00 175.00 180.00 185.00

Span Station

Ben

din

g M

om

ent

Aero Mesh BendingMoment

Beam BendingMoment

Refined FE BendingMoment

Summary of New Approach

• Costs– It approximates the inertial aeroelastic effect – It requires two spline models

• Benefits – It is computational (not heuristic)– It preserves full-vehicle balance– It is faster (process) and uses less CPU, too

• CPU savings accumulate as more Mach/Q pairs are used

– It creates fully reusable databases • easy to generate new trim states• able to recursively map the “mapped” solution to

other models