Applications of Advanced Composite Simulation and Design Optimization

Ming Zhou, Sam Patten, Martin Kemp, Robert Yancey, Erwan Mestres, Jean-Baptiste Mouillet
Altair Engineering, zhou@altair.com

1. Abstract
Usage of fiber reinforced composite material entered an new era when leading aircraft OEMs took an unprecedented step to design and manufacture essentially full composite airframe for commercial airliners. Composite structures offer unmatched design potential as the laminate material properties can be tailored almost continuously throughout the structure. However, this increased design freedom also brings new challenges for the design process and software. Moreover, as a relatively new material, composite behaviors are more complex and less fully understood by design engineers. Therefore, reliable simulation for highly complex events like bird strike and ditching can play an important role in shortening the product design cycle. This paper showcases two area of CAE tools for composite applications. On advanced simulation, bird strike simulation with Altair Radioss [1] is demonstrated on an aircraft underbelly fairing. On design optimization, an airplane wing structure is designed using an innovative composite optimization process implemented in Altair OptiStruct [1-3]. OptiStruct has seen increasing adoption among aerospace OEMs, as demonstrated in the Bombardier application process described in this paper.

2. Keywords: Composite Design, Composite Aircraft, Bird Strike Simulation, Composite Optimization. 3. Introduction
Computational tools have become an indispensible part of engineering tool set in today's product development environment. CAE software traditionally emphasize simulation of physical behavior of products in various operating conditions. In recent years, design optimization has become increasingly popular, largely driven by maturing software capabilities. This is a quite natural trend as optimization methods can help shortening the cycle of design improvements based on simulation results. In particular, concept generation through topology or Free-Size optimization brought game changing effect by inverting the order of CAD and CAE in the design process, bringing detailed performance consideration into the initial concept design stage. It has been shown that optimization technology can have significant impact on design innovation. For example, applying OptiStruct Airbus achieved over 40% weight reduction on a group of A380 leading edge ribs [5]. Similar successes have been demonstrated in all major industries including aerospace, automotive, heavy equipments, electronics, shipbuilding etc. This paper focuses on two aspects of advanced CAE tools for aerospace applications : (1) simulation of complex events such as bird strike using explicit finite element analysis; and (2) design optimization of composite structures. Composites behave far differently under impact than metals since they are made up of primarily brittle materials. Whereas metals deform and yield, composites tend to fracture. In many cases, the damage from impact occurs beneath the surface making damage difficult to detect visually. Composite damage mechanisms are also varied and can include delamination, fiber breakage, matrix failure or a combination of all of these. Damage progression is difficult to measure, even in laboratory settings, since the damage occurs internally and can progress very fast. Computational modeling of damage mechanisms in composites can yield significant information that can improve the design and performance of these materials. Models can predict damage progression, even at very small time scales, they can differentiate between different failure modes, and can handle the many variables involved in characterizing a composite structure. With an accurate damage model, many different design parameters can be analyzed, such as laminate lay-up, fiber volume fraction, and different combinations of fiber and matrix materials. With this ability, the best materials and designs can be chosen for specific structural applications. This paper presents an accurate simulation technique coupling the non linear explicit finite element simulation software RADIOSS [1] for modeling the impact of the bird on the structure
1

(impacted elements, impact forces, far field behavior) and DIGIMAT [4] for modeling the behavior of the composite material in the impact zone. The same software tools can also be applied to simulate ditching of an airplane in emergency water landing. The second part of the paper showcases application of an innovative composite design optimization technology implemented in OptiStruct [1]. Structural optimization has seen accelerated deployment throughout all industries in the past decade, largely due to the recognition that tremendous efficiency gain can be achieved at concept design stage through topology optimization. For metal structure, a two phase design process has become well established, where at Phase-I topology optimization is applied to generalize design concept, while design details are further optimized using sizing and shape optimization at Phase-II. For composite structure, the added design freedom prompts a modification of the process leading from concept to design details. While different forms of composite materials exist, the predominant usage is composite laminate where thin plies of various orientations are stacked together to form a shell structure. In recent years, the authors have developed a Three-Phase optimization process for composite laminate design optimization [2-3]. The target of the first phase is the material distribution in terms of orientation and thickness. This is achieved through free-size optimization where thickness of each 'super-ply' of a unique fiber direction is allowed to change freely throughout the structure. As a result thickness contour of each fiber orientation is obtained. A discrete interpretation of the thickness contour results in concept design of ply shapes and thicknesses. Then in Phase-II the interpreted ply-based structural model is further optimization under all design constraints with discrete design variables representing the number of plies of each ply patch. During Phase-III, ply stacking optimization is performed to refine the design according to detailed manufacturing constraints. It should be emphasized that manufacturing constraints are considered throughout all three optimization phases. For example, one important design/manufacturing requirement of aerospace OEMs is that no plies of same orientation is allowed to be stacked continuously for more than 3 or 4 plies. Such requirement would translate into percentage requirement during Phase-I and II so that a balanced distribution of fiber orientation is achieved to allow feasible stacking during Phase-III. The Three-Phase optimization process is illustrated in Fig. 1. This optimization process is implemented in the commercial software Altair OptiStruct. Another unique modeling technique developed in conjunction with the optimization process is a ply-based FEA model where PLY entities are defined as sets of elements [1]. Then ply layup is specified by a STACK definition. This modeling approach follows the design and manufacturing language known as Ply-Book. It provides a more intuitive interface between design and analysis, and makes design optimization more transparent. The software package OptiStruct has found quick adoption by OEMs producing aircraft, racing car and yacht made of advanced composites. Among these applications large to medium size airliners are of particular importance due to their enormous economic scale. In this paper a case study of a composite wing of a wide body long range aircraft is presented, followed by a description of the application process at Bombardier. Phase 1 Ply Tailoring?
Free Sizing!

Phase 2 Number of Plies? Phase 3 Laminate Stacking?

0

Rule based

Automation
Patch Interpretation

Ply Bundle Sizing!

45

-45

ply shuffling

90

45 -45 0 0 45 -45 90 90 -45 45 0 0 -45 45

Automation
Discrete Ply Thickness

Optimized Stacking Sequence!

Figure 1. Illustration of composite optimization phases

2

4. Bird strike simulation

RADIOSS is an explicit and implicit finite-element solver technology that simulates impact, mechanical, structural, fluid and fluid-structure interaction phenomena, taking into account nonlinear material, for quasi-static and dynamic loading events. RADIOSS is part of the Altair HyperWorks CAE technology suite [1]. It leverages a wide range of formulations such as Lagrangian, Eulerian, Arbitrary Lagrange-Euler (ALE) and Smooth Particle Hydrodynamics (SPH). RADIOSS accurately simulates the performance of structures that are subjected to large strains, displacements, and rotations by using a comprehensive library of material laws. DIGIMAT is a micromechanical modeling code that links with finite element analysis to determine the stress, strain, and damage at the constituent level. DIGIMAT [4] can be linked to RADIOSS through its user-defined material interface enabling the following two-scale approach: a classical finite element analysis is carried out at a macro scale, and for each time/load interval [t n , t n +1 ] and at each element integration point, DIGIMAT is called to perform an homogenization of the composite material under consideration (Figure 2). Based on the macroscopic strain tensor given by RADIOSS, DIGIMAT computes and returns, amongst other, the macroscopic stress tensor at the end of the time increment. The microstructure is not seen by RADIOSS but only by DIGIMAT, which considers, in the framework of a multi-scale approach, each integration point as the center of a representative volume element of the composite material. The material response computed by DIGIMAT will strongly depend on the phase behavior and the inclusion shape but also on the inclusion orientation. Another advantage of using DIGIMAT to simulate composite materials within FE analyses is that, in addition to the macro stress, DIGIMAT will compute stresses and strains in the phases and store it in the RADIOSS history variables. As described before, this is very useful, amongst other, in order to apply failure criteria at the microscopic level instead of the macroscopic level and to post-process these fields as any other macroscopic stress or strain fields. 4.1 Simulation R esults The application for this study was an underbelly fairing of an aircraft as shown in Figure 3. This part of the aircraft is subject to bird strike. The goal is to understand how a bird strike on the composite underbelly fairing will affect damage in the composite structure. The geometry of the fairing is shown in Figure 2. The idealized finite element representation of the underbelly fairing was generated using HyperMesh. The loading on the model represented 2 equipment masses, weighing 2Kg and 3Kg each, which were connected to their mounting locations through the use of RBE3 elements. Altairs RADIOSS finite element solver has a validated bird model based on Smooth Particle Hydrodynamics (SPH) theory that was used to simulate the effect of a bird strike on the fairing. The model has been validated with extensive test data. The model was run in Radioss with the stress, strain, kinetic energy, velocity, deceleration information, and deformation limits predicted. The fairing is modeling as a 0/90 weave material. The material properties of the fibers and matrix materials are taken from Kueh and Pellegrino [6] and are as shown in Table 1. An M18 epoxy matrix and T300 carbon fiber are modeled. The matrix is isotropic and the fibers are transversely isotropic. The fiber volume fraction is 60% and the percentage of 0 and 90 layers is 50% each. A DIGIMAT material model is developed with these parameters yielding a composite longitudinal and transverse modulus of 75602 MPa (equal number of 0 and 90 plies) and a composite density of 1460 kg/m3. The Radioss SPH bird model is then run on the structure with a velocity of 180 mm/ms and angle of attack of 30 degrees to the fairing surface. The bird model has been validated against significant test data. The results are shown in Figures 4-6. Figure 4 shows the macroscopic 11 stresses and 11 strains (along the X-axis). This is what is normally predicted for a finite element analysis. Damage occurs at the micromechanical level so it is important to be able to decompose the results into fiber and matrix stresses and strains. Figure 5 shows the 11 strain and first principal strain in the epoxy matrix and Figure 6 shows the first principal strain in the 0 fibers and the first principal strain in the 90 fibers.

Figure 2: Interaction between DIGIMAT and RADIOSS. Left: Classical FE procedure Right: Multi-scale procedure using DIGIMAT as the material modeler.

Figure 3. Depiction of an aircraft underbelly fairing

Figure 4. Geometry model of the underbelly fairing

4

Table 1. Material Properties Property Density Longitudinal Youngs Modulus (E11) Longitudinal Poissons Ratio (11) Transverse Youngs Modulus (E22) Transverse Poissons Ratio (22) Transverse Shear Modulus Ultimate tensile Strength Ultimate compressive Strength Ultimate tensile strain Ultimate compressive strain Matrix (M18) 1160 kg/m3 3500 MPa 0.38 3500 MPa 0.38 Fiber (T300) 1760 kg/m3 233000 MPa 0.2 23100 MPa 0.2 8963 MPa 1990 MPa 1990 MPa

0.037 0.0481

11 (X-Direction) Macroscopic Strain

11 (X-Direction) Macroscopic Stress

Figure 5. Macroscopic stress and strain in composite structure under bird strike loading condition

11 (X-direction) Strain in Matrix

First Principal Strain in Matrix

Figure 6. Computed strains in M18 Epoxy Matrix

First Principal Strain in Longitudinal (0) Fibers

First Principal Strain in Transverse (90) Fibers

Figure 6. Computed First Principal Strains in Longitudinal and Transverse Fibers When introducing failure criteria in the DIGIMAT material model, the mechanism of microscopic failure can be modeled. Different failure indicators have been introduced, acting at the matrix level or at the fibers level. For the Epoxy matrix, the failure indicator consists of a maximum principal strain value of 0.037 in tensile direction and of 0.0481 in compression. For the Carbon fibers (both for longitudinal and transverse fibers), the failure indicator consists of a maximum principal stress value of 1990MPa. Consequently, an element can be deleted once one of these failure indicators is reached. It is then possible to better understand the failure mechanism. Figure 7 below illustrates the damage at the macroscopic and at the phase levels at time 6.2s. A value greater than 1 (shown in red) for a failure indicator means the layer has failed. For the comparison, a macroscopic Tsai-Wu failure indicator has also been used. The Tsai-Wu criterion shows more damage than is shown in the phase level damage plots. Tsai-Wu is generally considered a fairly conservative damage model. Being able to predict damage at the micromechanical level will lead to more accurate damage predictions which will probably result in lighter structures. Also, being able to predict fiber vs. matrix failure will lead to better insights about how to design robust yet lightweight designs. 4.2 C onclusions In this paper, it has been shown that the behavior of a composite has to be considered at the micromechanical level. At first, a DIGIMAT material model allows obtaining the strain and stress tensors at the matrix and fiber level. Moreover, failure indicators can be used not only at the macroscopic level, but also at the phase level. Failure can occur because of the fracture of the matrix, as well as of the carbon fibers or of the macroscopic structure. The considered DIGIMAT model allows combining all these failure indicators to obtain an accurate modeling of the failure mechanism. Future work will involve investigating additional laminate configurations and improving the overall efficiency of the process. Extensive test evaluation has been conducted but more will be required to develop robust analysis processes for impact simulation of composites.

Figure 7. Failure indicators in Carbon fibers, in Epoxy matrix phase and at macroscopic level.

5. Airplane wing optimization

The Three-Phase composite design process is demonstrated through the design of composite wing covers of a wide body aircraft shown in Fig. 8. Ten load cases of key significance are considered. In this simplified concept proof exercise only constraints on wing tip displacement and rotation are considered, with upper bounds not exceeding those of a baseline aluminum wing under each load case. Only carbon fiber composite top and bottom covers are optimized. Ply orientations available are 0, +45/-45, 90 plies, with leading edge as reference. 5.1 Phase-I: Concept Design - Free-Size Optimization Manufacturing Constraints considered include: max. thickness of each fiber orientation 10 mm +45/-45 plies balanced 8 mm total laminate thickness 32 mm min. percentage of an available fiber orientation 10% The thickness distribution of the four fiber orientation is shown in Fig. 9. The results show that the 0 ply requires the highest number of layers. The 45 & -45 layers are identical, meeting the manufacturing requirement, and show that a number of 45 /-45 layers are required in the centre of the covers. The 90 ply requires very few layers as expected. The thickness is maintained at leading & trailing edge of the wing centralized region, providing torsional stiffness as well as bending stiffness. The thickness is reduced in other regions to minimize the mass.
7

Figure 8. Composite wing model of a wide body aircraft

Figure 9. Thickness distribution of the upper and lower covers of the wing (0 left; 45/-45 middle; 90 right) At the end of the Free Element Sizing optimization, OptiStruct automatically generates ply shapes based upon the optimization results. For each ply orientation, the optimized ply thickness is split into a number of layers of different shapes, the default being 4 shapes per orientation. Fig. 10 shows ply shapes automatically generated by OptiStruct.

Figure 10. Automatically generated 0 ply shapes of the upper cover The ply shapes that were generated by OptiStruct were edited using HyperMesh. This allows the shape of the plies to be based around the optimization results but also be made smooth and manufacturable. Any infeasible ply shapes can be removed and shapes which are too complicated can be simplified. This process was completed for each of the layers that were generated automatically, producing a total of 62 ply shapes. Fig. 11 illustrate the process of such manual editing. These plies will be used in the system level optimization at Phase II when determining how many plies are needed.
8

Figure 11. Ply Shapes Modified Using HyperMesh 5.2 Phase II: Design Fine Tuning Ply-Bundle Size Optimization The sizing optimization problem remained the same as in phase-I for simplicity in his concept proof study. For real applications this optimization stage should consider all design criteria such as strength and stability constraints. The total thickness contour after sizing optimization is shown in Fig. 12. Of the 62 ply shapes available in the study, 59 have been retained. The other 3 have been automatically removed by having their thicknesses tuned to zero.

Figure 12. Total thickness contour of upper and lower covers after size optimization After completing the optimization, OptiStruct automatically creates physical plies for the detailed design stage of the optimization. This is achieved by dividing the optimized thickness by the discrete step size (thickness of a single ply) to determine the number of physical plies to create. Note that size optimization can also be carried out with discrete ply thickness directly. The size optimization automatically generated a new model which contained a total 341 ply layers as follows: 174 plies with 0 orientation 57 plies with 45 orientation 57 plies with -45 orientation 53 plies with 90 orientation
9

The model now has optimum shaped plies and also the number of each ply has been tuned. To complete the process, it is necessary to also meet the ply lay-up rules. This is addressed in the next phase of the design. 5.3 Phase III: Ply-Book Details Ply Stacking Sequence Optimization This optimization phase focuses on detailed finish of the final ply-book while preserving both manufacturing and performance constraints. Additionally, it is required that certain ply book rules be applied to guide the stacking of plies based on specific requirements. Some of the ply book rules that control the stacking sequence are: Maximum number of successive plies of a particular fiber orientation Pairing of the + and - 45s Identifying a sequence for the core and cover regions For this example, the optimization problem as previously formulated in the sizing phase is retained and the following additional ply book rules are applied: (a) the maximum successive number of plies does not exceed 3 plies; (b) the + and - 45s be reversed paired. Fig. 13 shows a HTML report of the stacking sequence evolution of the upper cover, generated automatically by OptiStruct. It shows some general trends: (1) 0 plies moved towards the outside of the covers; (2) 90 & -45 plies moved towards centre; (3) 45 plies distributed to break up large ply groups.

Figure 13. Stacking optimization - HTML report of stacking evolution of the upper cover For a more detailed review of the layup, a utility is available in HyperMesh to show solid view for visualization purposes. The utility allows thickness scaling and also applies the element & laminate shell reference plane offsets.

10

Figure 14. Bottom Wing Cover Solid Visualization of Plies (Thickness x5) The solid visualization makes it possible to see: (1) How the plies transition / overlap across different zones; (2) Thickness drop off; (3) Manufacturing sequence (ply book). Also when using a solid visualization in conjunction with the cross sectioning tools in HyperView, it is possible to perform a detailed investigation into the layup throughout the length & width of the covers. Another benefit would be for generating an inner mould surface for the laminates. 5.4 Conclusions Through this study, the Three-Phase optimization process has successfully demonstrated its capacity for maximizing utilization of the potential of composite material, while significantly shortening the design process. As a simplified concept proof study detailed strength and buckling criteria necessary for design phases II and III are ignored. However, these constraints can be efficiently addressed in the optimization process, as shown in a Bombardier application discussed in the next section.

6. Application of Altair composite design optimization process to aero-structure composite component development at Bombardier
This section outlines application of the Altair Composite Optimization technology to composite component design at Bombardier. As part of Bombardiers ongoing technology development initiatives, application of the process was explored at single and multiple component levels. A description of the process and method of application inside a dynamic aerospace design environment is described. Methods for incorporating structural and manufacturing constraints are introduced. The interfaces developed between design and stressing groups are also summarized, which underpin the successful application of the technology in an environment where design requirements can frequently change. 6.1 Integration of Altairs Composite Design Process Integration of Altairs composite optimization process with the design process and all of the necessary interfaces is shown schematically in Fig. 15. The main additions to the process are interfaces accommodating inputs and outputs to and from the design team. Notably, custom responses and constraints are needed to align the optimization with strength, stiffness and stability qualification requirements. Export of the optimization solution is also required in a number of different formats including: CAD format laminate descriptions, qualification report summaries and additional finite element formats.
11

Bombardier FEM
Concept
Optimization Model Setup Freeform optimization (zonal boundaries) Efficient Sensitivity Studies Custom Qualification Constraints (Strength, Stability)

Altair Composite Optimization Process

Fine Tuning

Ply Continuity Locked in Manufacturing Rules Additional Custom Qualification Constraints

Delivery

CAD format laminate definitions Finite Element format conversion Qualification reporting

Bombardier Design

Figure 15. Schematic Summary of the Integration of Altairs Composite Design Optimization Process with Bombardiers Aero-Structure Design Process 6.2 Composite Optimization Interfaces A review of the Bombardier Aero-Structure design process was performed to identify the inputs and outputs required for the composite optimization process. Successful access to the technology in the overall design process is underpinned by these interfaces working efficiently and robustly. The main focus areas for the interface development were: i) Conversion of Bombardier FEM data to OptiStruct format suitable for optimization ii) FEM export at the end of the process iii) CAD format export of final designs iv) Qualification Analysis reporting in Bombardier format (Spreadsheets and other digital documents) Altairs generic FEM and composite interfaces were modified to facilitate each of these requirements in the Bombardier design environment. The resulting solution was a single integrated platform which facilitated passage of input and output data to and from the optimization between Bombardier and Altair. Composite specific results visualization and report data could easily be shared and reviewed by all parties. 6.3 Optimization Problem Formulation The optimization problems were typically defined to minimize mass subject to stiffness, allowable composite stresses and stability criteria. Multiple load cases were defined and, where available, appropriate stiffness targets set for each based on the baseline response. In addition to the composite laminate sizing design variables for components, shape optimization of the stiffening members was also investigated through Free Shape technology available in OptiStruct. Greater design freedom is afforded with this approach as it allows each stiffener height to change independently and freely in shape as well as size. This is often advantageous where a balance between relative stiffness and stability must be maintained. To constrain the optimization to derive designs compatible with the design team requirements for some components, zone boundaries were defined over the surface. OptiStruct can constrain the laminate solutions to
12

respect these boundaries from the first free sizing stage. This is often a key manufacturability requirement and can be locked down at the concept stage. Commonality between manufacturing constraints was maintained throughout the stages to enforce minimum percentages of cloths and uni-directional plies in the stack. In the later stages, manufacturing rules were enforced limiting the maximum number of consecutive plies. The structural constraints were implemented by direct sampling of finite element results (stiffness and strain) or by custom calculations developed to correlate with qualification assessment methods (global and local stability, additional strength requirements). The custom calculations were implemented through OptiStructs DRESP3 functionality, which ensures efficiency in the handling of custom calculation routines and response sensitivities. 6.4 Discussion The composite optimization process was applied successfully in a real world aerospace design environment allowing efficient exploration of designs and delivering weight saving potential for a range of components and systems. The following major advantages were found from application of the process: i) The free-form stage provided an efficient testing ground for design sensitivity to applied loads and design constraints. The solutions were not influenced by previous designs and provided insight of methods for improving structural efficiency. This provides a very efficient method for performing trade off studies and rapid assessment of changes in design requirements. ii) The process demonstrated the value of locking ply continuity into the optimization from early in the process. In this way, manufacturability could be constrained with less impact on the structural efficiency. Interfaces were developed between the OptiStruct ply based output and design system carrying over ply continuity directly. iii) Significant mass savings were predicted from application of the technology and a measure of the effect on weight of varying manufacturing constraints could be quantified. iv) The input data and optimization solutions could be integrated with the current design practice at Bombardier, facilitating efficient communication and final design qualification. Application of the optimization approach at Bombardier has led to a repeatable process, which accommodates the composite design qualification requirements and can be enhanced and applied at component and system level.

7. Conclusion
Increased composite usage for new generation airliners brought new opportunities and challenges for CAE tools. This paper highlights two important areas of evolving technologies: (1) Simulation of highly complex events such as bird strike and ditching; (2) Optimization for efficient composite design. Both can bring significant value for shortening airplane design cycle and enhancing structural performance/weight ratio. In the first part Bird strike analysis using Radioss explicit finite element analysis combined with DigiMat multi-scale material model is demonstrated on an airplane underbelly fairing. The second part focused on application of an innovative optimization process of OptiStruct. A aircraft wing case study is shown to demonstrate the process that leads the design from ply shape concept generation to final ply-book details. Then a detailed description of the application within a real world aircraft design environment at Bombardier Aerospace is given. It is particularly notable that customer specific design constraints on panel strength and stability are incorporated through external responses DRESP3. This demonstrated the versatility of the software package that allows the optimization process to fit into an established complex environment of commercial aircraft design.

8. Acknowledgement
The authors are indebted to Bombardier Aerospace for their permission to publish the application process of Altair composite optimization process within their aircraft design environment. 9. References [1] HyperWorks Software, Altair Engineering, Troy, MI.
13

[2] Zhou, M., Fluey, R., Willmet, T.: Multiple phase optimization of composite structures. Proc. 9th US National Congress of Computational Mechanics, San Francisco, July 2007. [3] Zhou, M., Fluery, R., Dias, W.: Composite design optimization from concept to ply-book details. Proc. 8th World Congress of Structural and Multidisciplinary Optimization, Lisbon, Portugal, June 2009. [4] DIGIMAT Software, e-Xstream engineering, Louvain-la-Neuve, Belgium. [5] Krog, L., Tucker, A., Kempt, M., Boyd, R., Topology Optimization Of Aircraft Wing Box Ribs, AIAA-2004-4481, Proc. 10th AIAA/ISSMO Symposium on Multidisciplinary Analysis and Optimization, Albany, NY, 2004. [6] Kueh, A. and S. Pellegrino (2007). ABD matrix of single-ply triaxial weave fabric composites. 48th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, 2007, Honolulu, Hawaii.

14