Bird Strike Simulation Using ANSYS LS/DYNA Carlos Shultz Jim Peters Phoenix Analysis And Design Technologies, Inc.
Abstract Phoenix Analysis and Design Technologies (PADT) has developed a methodology for bird ingestion simulation (bird strike) of aircraft engine fans which reduces the time to prepare, solve, and post-process to less than 2 weeks for initial iterations and less than 2 days for subsequent iterations. The method uses shell elements for the blades, rigid solid elements for the hub, shell elements for the shroud, and solid ALE (Arbitrary-Lagrangian-Eulerian) elements for the Euler mesh to transport the bird material.
Introduction Bird strike analysis is a common type of analysis done during the design and analysis of aircraft engine fans. These simulations are done in order to predict whether various designs designs will pass the necessary certification tests. For this type of simulation to be useful in a design environment, the time to perform the simulation, measured from delivery of a new blade to the reporting of predicted plastic strains, must be on par with the time spent by the other engineers analyzing the same blade for their disciplines (aerodynamics, mechanical structures, heat transfer, and manufacturing). Typically, the physics modeled during a bird strike simulation include bird material contacting the blades, bird material contacting the shroud, and blades contacting the shroud. This paper discusses the steps for quick simulation of a bird strike which captures all of the necessary physics to accurately predict how a fan will react to a bird ingestion.
Finite Element Modeling Preparing Models (*.db) Figure 1 shows a sector of a typical mesh used. Note that each element of the blade has a different real constant due to the variations of thickness. thickness. Figure 2 shows a cross-section view of a typically model. Each of the components shown in Figure 2, are discussed briefly below.
Blades (airfoils) Airfoil meshes are created in ANSYS with Shell163 elements elements (Belytschko-Tsay formulation). Macros are typically used to create shell elements from 1 and 2 element-through-the-thickness solid models used in static structural analyses. The thickness at each node is defined appropriately for a smooth mesh transition. Mesh density should be biased to provide refinement at the expected point of contact.
Hub The hub mesh is created as rigid Solid164 elements. elements. The hub mesh shares nodes nodes with the blade mesh at the root of the airfoils. Occasionally the rigid elements will extend into the airfoil (depending upon the aspect ratio of the shells at the root of the airfoil).
Shroud The shroud mesh is created with Shell163 elements (Belytschko-Tsay formulation). Boundary conditions are applied to hold the shroud in place. Shroud mesh density should be similar to the mesh density of the blade at the tip.
Bird Material The bird material is simulated with an Euler mesh. The Euler mesh is a stationary mesh which transports the bird material within it. The material occupies a volume fraction of each element. At the beginning of the analysis the initial bird material elements have a fraction equal to 100% and the rest of the elements equal 0%. As the analysis progresses, the bird material translates through the mesh changing the volume fractions of the elements which contain it. This means that when a shell element (Lagrangian Mesh) from the airfoils sweeps through a brick element (Euler Mesh) of the air that has a volume fraction greater than a threshold value, an additional set of calculations is triggered which is the mechanism of momentum transfer between the bird material and the airfoils.
Parts Meshing within ANSYS must be done with care to minimize effort in subsequent steps when editing the *.k file to modify material models, element types, and other part specific commands. A model with several hundred PARTS is typical because each element in the airfoil has a different real constant set of thicknesses. Part numbers are normally assigned to the mesh from lowest to highest node/element number. One way to assign parts in the proper order is to use the EDPART command on selected parts of the model in the following manner. •
Select elements to be Part 1
•
EDPART, create will create Part 1
•
Also select elements to be Part 2
•
EDPART, update will create Part 2 without changing Part 1 even if the numbering of the nodes/elements of Part 1 are greater than Part 2. Note, if instead EDPART was issued, the Parts would order by their node/element numbering.
By careful management of the parts list, one can easily re-mesh intermediate parts in your parts list without having to change the manual editing in subsequent steps.
Boundary Conditions An important part of this type of analysis is the initial and final conditions of the model. Loading the fan with the proper rotational velocity prior to the bird strike event is a crucial aspect of the analysis. Many different techniques exist to do this. Good success has been achieved by doing the ramp-up entirely within LS-DYNA. Monitoring the blade tip velocity (relative to the hub) is a good way to determine whether the vibrations due to startup have subsided. Once the blade has ramped up to speed, the bird can be set into motion. Not to be overlooked, ramp down has some significance as well. Most post-processing involves visualization of the plastic strains and displacements, which are easier to do when the model is at rest. Ramp down can be done at the same rate as ramp up but must also include a period of rest at zero velocity for the blades to come to rest.
Contact Contact (and friction) is often used to transfer loads between the blades and external structures (shroud). When frictional contact is active between the blades and external structure, the time step drops by a factor of 10 to 100.
Create the LS/DYNA input File Create the *.k file. The *.k file is a text file containing all the model information in a defined fixed field format. Details regarding the commands within the *.k file can be found within the LS/DYNA documentation available from LSTC at: http://www.lstc.com/pages/manuals/index.html. ANSYS users note: LSTC may be contacted for documentation, but Technical Support must be obtained by your local ANSYS Support Distributor.
Preparing Models (*.k) Once the *.k file is written, edit the necessary cards to make the changes described below. Some of these changes are done manually because ANSYS doesn’t yet support the feature; other changes are done because it is easier to make the changes manually.
Materials High strain rate events require high strain rate material models. LS/DYNA supports a large variety of material models. For metals a popular choice is to use *MAT_PIECEWISE_LINEAR_PLASTICITY . For bird material use *MAT_NULL and *EOS_TABULATED corresponding to 90% water and 10% air. It is often easier to create the *.k file with placeholder material models which are replaced by editing the file manually.
Hourglass Hourglass energy is often a problem with the reduced element formulations used in explicit analysis codes. Controlling the hourglass energy is done by using *CONTROL_HOURGLASS or *HOURGLASS . Option 2 (default viscous hourglass control ) is typically used for this type of analysis,. Viscous control is recommended for high velocity deformations.
Damping Damping of unwanted vibration is another common problem encountered. Good success has been obtained using *DAMPING_RELATIVE to stabilize the blades during ramp-up and ramp-down. This card provides blade damping proportional to the relative velocity of the blades and the hub. Create the necessary part sets and add this command to the *.k file.
ALE (Arbitrary-Lagrangian-Eulerian) There are several changes that need to be made to the *.k file to support the ALE formulations used for the bird material mesh. The *SECTION_SOLID cards for the parts associated with the Euler mesh need to be changed to *SECTION_SOLID_ALE option 12. The part associated with the Euler mesh not initially occupied by the bird material needs to have *INITIAL_VOID_PART applied to it. This card initializes that mesh to have 0% volume fraction. The control of the Euler mesh is done with *CONTROL_ALE card. Use of *CONTROL_ALE with a currently undocumented advection control is typically used.
The interaction between the Euler mesh and the Lagrangian mesh is controlled by the *CONSTRAINED_LAGRANGE_IN_SOLID card. This card is used on part sets which are generated using the *SET_PART_LIST card.
Solving Models (*.k) Once the *.k file has been modified, the model is ready to solve. Setting the MEMORY option is often necessary for large models. Use Version 9.60, which is included with the distribution of ANSYS6.0, since several of the new features added to that version of LS/DYNA are required for this analysis. The timestep for models as described above are generally in the range of 1e-6sec. This size of timestep allows a complete analysis to run within 2-4 hours (if a shroud is not included) and 1-2 days (with a shroud) on a Pentium III 1GHz, 1 Gb RAM, and IDE hard drive. Running a similar analysis with solid elements for the blades requires a timestep in the range of 5e-8sec. Since run times vary inversely to timestep size, solid blade models tend to take too much time to solve.
Post Processing Models (d3plot) The bulk of post-processing is done within LSPOST, the post-processor included with the distribution of LS/DYNA that comes with ANSYS. Typically plots of deflections and plastic strain in the airfoils are of interest. Animation of the volume fraction of the Euler mesh also shows valuable information. Figures 3, 4, 5, and 6 are plots from LSPOST showing the volume fraction of bird material at various times throughout a demonstration simulation. Figure 3 shows the initial state of the fan and 2 birds at rest. Figure 4 shows the 2 birds in motion. Figure 5 shows the 1st impact as a blade hits the 1 st bird. Figure 6 nd shows a subsequent impact as a blade hits the 2 bird. Figures 7 and 8 are plots from LSPOST showing the plastic strain and effective (von mises) stress in the blades at time = 0.085094 sec which is the same time frame as figure 6.
Post Processing Models (*.rst) Another common processing technique is to export the nodal coordinates from LSPOST and then put them into ANSYS results using dnsol. This allows users to create an *.rst file and use established macros for post processing. Figures 9 and 10 are plots from ANSYS showing the final position of the blade, it is rotated to the correct location to show the final blade shape with respect to the original shape. The deformed mesh can then be used to calculate various pass/fail criteria.
Future Work There can never be enough model correlation. Every family of fans is different and requires some amount of correlation to real test data. Eventually a set of correlation parameters could be developed with a standard range of recommended values. Simulations in which additional engine structures, like downstream vanes, included would be interesting to investigate.
Conclusion Simulations predicting the plastic strains resulting from a bird ingestion simulation (bird strike) of aircraft engine fans can be performed fast enough such that the results can be incorporated into the normal design iteration process.
References 1.
Olovsson, Lars; Souli, Mhamed; Do, Ian, LS-DYNA – ALE Capabilities (Arbitrary-Lagrangian Eulerian) Fluid-Structure Iteration Modeling, LSTC, 2002
2.
Iannucci, L, Bird-Strike impact modeling , ImechE (1998), S531/002/98
3.
Budgey, Richard, The Development of a Substitute Artificial Bird by the International Birdstrike Research Group for use in Aircraft Component Testing , IBSC25/WP-IE3, April 2000
Figures Figure 1. Sector of a fan with an airfoil (shells) and a hub (solids), the mesh is connected at their interface and mesh is continuous from sector to sector
Figure 2. Section plot of a typical bird strike model (without a shroud)
Figure 3. LSPOST plot of the volume fraction at time=0
Figure 4. LSPOST plot of the volume fraction at time=0.037997, both birds in motion
st
Figure 5. LSPOST plot of the volume fraction at time=0.059398, 1 bird has hit the fan
nd
Figure 6. LSPOST plot of the volume fraction at time=0.085094, 2
bird has hit the fan
Figure 7. LSPOST plot of the plastic strain at time=0.085094, same time frame as Figure 6
Figure 8. LSPOST plot of the effective stress at time=0.085094, same time frame as Figure 6
Figure 9. ANSYS displacement plot of a fan blade at rest (pldisp,2)
Figure 10. ANSYS displacement plot of a fan blade at rest (plnsol,u,sum)
