OPTIMIZATION ASSISTED STRUCTURAL DESIGN OF THE REAR FUSELAGE OF THE A400M, A NEW MILITARY TRANSPORT AIRCRAFT Gruber, H.*; Schuhmacher, G.***; Förtsch, C.**; Rieder, E.* *Altair Engineering, München **Altair Engineering, Böblingen ***EADS Military Aircraft, München Summary: The following paper shows how topology optimization methods have been used in order to find new concepts for a complete structure. The first part gives an overview of the functionality of the optimization software Altair OptiStruct, and points out the key features used for this project. In the second part, an example for the usage of topology optimization on system level, for the complete rear fuselage of A400M transport aircraft, is shown. Keywords: topology optimization, system level, optimization software NAFEMS Seminar: „Optimization in Structural Mechanics“ 1 April 27 - 28, 2005 Wiesbaden, Germany 1 Introduction In the past years Topology optimization methods have been included into standard development processes of the automotive and aerospace industry. Innovative CAE-Systems offer many possibilities to establish weight optimized or load path optimum designs to reduce costly and time consuming iterations in the development process. Examples of applications for these methods are mostly single components, like wheel trunks. In this paper topology optimization has been applied at system level, in order to determine a new concept of the complete rear fuselage. The main goal of this project was to establish new ways in designing the complete rear fuselage. Explained in simple words, the rear fuselage of the A400M is a tube with a big cut which is loaded predominantly with torsion. Designing this part of the A400M as a lightweight structure is a special challenge. In a first step, the objective was to find the optimum material distribution with respect to a given design space. The design space is bounded by the outer loft, driven by aerodynamic requirements and the inner loft determined by the space required for passengers, cargo load, systems and other functional requirements. The results were interpreted into an aerospace-like structure, consisting of stringer and frames which was the input for another optimization step. As a result a detailed design proposal is obtained which is important to make established design decisions. These optimization steps lead to a significant weight reduction compared to the baseline design and showed new ways to fulfil different requirements. 2 Optimization software 2.1 Formulation of the optimization problem Most of the software used for structural optimization is based on finite element methods (FEM). First used methods are size optimization, here for example shell thicknesses or beam parameter are the design variables and shape optimization in which node coordinates of FEM meshes are parameterised in order to vary the shape of an existing component. Both methods are based on a fix topology. At the end of the 90’s, methods were introduced to find an optimum layout of a structure which was called topology optimization. Based on a design space, modelled with finite elements and the relevant load cases, an optimum material distribution can be found. All these methods work iterative and therefore require a series of FE analyses to an optimal design (Fig. 2) [3]. In structural optimization the formulation of the optimization problem is important. It can be formulated as follows: Minimize : f(x) = f ( x1, x2 ,K, xn ) Subject to : g j (x) ≤ 0 j = 1,K, m hk ( x) = 0 k = 1,K, mh xiL ≤ xi ≤ xiU i = 1,K, n Fig. 1: General formulation of the optimization problem The so called optimization model consists of the objective function f, the design constrains g and h and the design variables x. 2.2 Mathematical approach The optimization software Altair OptiStruct used in this project has a general mathematical approach. The mechanical optimization problem is transferred into a mathematical one and then solved by the optimizer. Compared to software working with optimum criteria there are following advantages [1]: − − − − The user has maximum freedom in formulating the optimization problem In principle all results of a FEM Analysis are available as a response for optimization Responses of external servers could be used, e.g. damage values from fatigue Different methods for optimization (chapter 2.1) could be combined in one run NAFEMS Seminar: „Optimization in Structural Mechanics“ 2 April 27 - 28, 2005 Wiesbaden, Germany First step in an optimization run is the finite element analysis which is necessary for the sensitivity calculation. Based on this the search direction for the optimization algorithm is found, which leads to a variation of the design variables to run this loop again with a better draft. This iteration process stops when stop criterion is reached. Fig. 2: Scheme of an optimization run Altair OptiStruct uses an integrated FE-Solver and so sensitivities, the derivatives of a response g with respect to the design variables x can be calculated very cheaply because the factorised stiffness matrix K can be used and only a forward-backward division is necessary [5]. g j ( X ) = Q Tj U , KU = P (I) ⇒ ∂g j = ∂x i ∂Q Tj ∂x i U + Q Tj ∂U ∂U ∂P ∂K Uj , K = − ∂x i ∂x i ∂x i ∂x i g j ( X ) = Q Tj U , KU = P (II) ⇒ ∂g j = ∂x i ∂P ∂K U + U jT − U with KU j = Q j ∂x i ∂ x ∂ x i i ∂Q Tj Equation (I) shows the direct differences method which is efficient when less responses than design variables are used, g << x (size and shape optimization). The adjoint method (equation II) is preferable, if the number of responses is greater than the number of design variables, g >> x (topology optimization). Based on the sensitivities the optimization problem is approximated. The behaviour of the design constrains and objective function are considered in a small proximity around the current draft. Thus the total number of iterations and expensive FE-Analysis runs are reduced [3]. NAFEMS Seminar: „Optimization in Structural Mechanics“ 3 April 27 - 28, 2005 Wiesbaden, Germany ∂g g~ j (x) = g j 0 + ∑ j ( xi − xi 0 ) i ∂x i (III) Equation III shows a linear approximation of a response g against the gradient. In addition reciprocal and convex approximation types are also used in OptiStruct. 2.3 Topology optimization 2.3.1 Density method (SIMP) That approach for topology optimization (solid isotropic material with penalty) is based on a design space meshed of finite elements. Every element represents one design variable in the optimization problem which leads to a variation of the stiffness and the density. Due to exponent p, the coherence between a normalized density ρ/ρ0 and the young modulus E shows a hyperbolic behaviour. Resultant from this approach, semi dense elements should be penalized to decide whether to be full material or a hole. At the end the design result should be discrete which is easier to be interpreted. [2]. Fig. 3: Approach of the density method Figure 3 shows a result of this approach. On the black contoured elements material is needed and for the white elements material can be dismissed for an optimum structure. 2.3.2 Integration of manufacturing constrains Not in every case is the optimum material distribution the best in developing a component. Manufacturing aspects have also to be considered in topology optimization, to ensure that design proposals from the optimizer can be transferred into a construction without fundamental changes. To include these aspects into the optimization process, special features can be activated by the user in order to control the design result. Fig. 4 shows such a manufacturing constrain (minimum member size control), which suppresses small design features and create bigger structures – not smaller than a diameter d - instead. Switching on a manufacturing constrain has more positive affects: − − − Reaching a discrete material distribution Checkerboard control Avoiding mesh dependency [5] NAFEMS Seminar: „Optimization in Structural Mechanics“ 4 April 27 - 28, 2005 Wiesbaden, Germany Fig. 4: Effect of manufacturing constrains on the design result More manufacturing constrains which are available in Altair OptiStruct are: − − − − Maximum member size (spitting up material accumulations) Draw direction Symmetry Pattern repetition [1] 3 Concept studies using topology optimization considering the A400M airfreighter tail structure as an example 3.1 Requirements on an ideal tail structure The A400M’s tail structure consists of an outer skin, supported by stringers and vertical frames. This tube has a big tailboard for loading and unloading goods or even vehicles. It can be closed with a ramp and a door both outwards and inwards. Section C Section B Section A Cargo Door Ramp Longerons Floor Fig. 5: Baseline design (half model) The objective of the optimization is to increase the stiffness while keeping the mass bounded above. As the door stands slant under tensional load, the relative displacements shouldn’t exceed a certain value to allow closing the door securely during the flight after launching cargo or paratroops. Particularly critical are load cases caused by extreme manoeuvres or squalls which load the fuselage by torsion or bending. NAFEMS Seminar: „Optimization in Structural Mechanics“ 5 April 27 - 28, 2005 Wiesbaden, Germany 3.2 Design space and optimization model In this case the maximum allowable design space is limited by the outer skin of the fuselage and inside by system requirements, e.g. space for opening and closing door and ramp, loading space, etc. . Within this area the material can be distributed arbitrarily during the optimization to replace the stringers and vertical frames, which were removed from the original model, in order to increase the total stiffness of the afterbody. Various other components, e.g. door and ramp, serve during the calculation as non design, they are considered to be unchangeable. Beneath the topological design space the thicknesses of the main fields were defined as variables to enable the Optimizer to decide weather the load path should be carried by the outer skin (via increasing the thickness) or rather by using additional braces inwards. Sec. B: Frames removed Sec. C: Frames removed Sec. A: Frames removed Design Space modelled by Solids + Floor & longerons removed Fig. 6: Non design and design space The optimization task includes the minimization of the linear combination of the Compliance, an equivalent of the extension energy for all relevant loading cases (k). As design constrains for the calculation, the exceeding of the mass – in comparison to the original model – and the relative displacement between door and fuselage structure are limited. This conforms to a modified matter of formulating the traditional topology optimization task. Minimize { f ( ρ,t ) = ∑ f kT u k } with k p = topology variable Subject to m ( ρ,t ) ≤ m baseline t = thickness variable d 1 ( ρ,t ) − d 2 ( ρ,t ) ≤ d max ′ ≤ ρi ≤ 1, i = 1,...,n ρmin fTu = compliance m = mass d = displacement t min ≤ t j ≤ t max , j = 1,...,m Fig. 7: Formulation of the optimization task 3.3 Design result of the 3-D Topology optimization The result of the first optimization step is a qualitative material distribution within the design space idealized of 3-D volume elements, as well as a set of thickness variables for the main fields. Figure 8 shows this distribution, where all non design areas and the outer skin are blanked. It shows the accumulation of material in the area of the load introduction underneath the tail unit, along the original edge beams (longerons) and confirms the importance of the original model’s details. Nevertheless the result shows new insights and design elements, e.g. an extended inner skin in the area of the door as well as an intelligent connection of the door and the fuselage structure. NAFEMS Seminar: „Optimization in Structural Mechanics“ 6 April 27 - 28, 2005 Wiesbaden, Germany Fig. 8: Material distribution regarding topology optimization on volume meshed design space A kind of kinematics like scissors takes care that the side conditions limitation of the relative displacement between door and fuselage structure can be fulfilled. This is an example that a complex formulation of an optimization problem can lead to smart ideas and solution, which are hard or even impossible to obtain only by intuition. Fig. 9: Smart design proposal for a connection cargo door to the substructure 3.4 Results of the continuative topology optimization on shell base Starting from the abovementioned 3-D material distribution, another design space model meshed with shell elements is originated. This continuative step is performed to make a detailed statement about the designs of the single components and determine the ideal stiffness distribution for the complete tail structure. The optimization model is not changed in comparison to step one. For the visualization of the design results for this case, a contour display of the element density suited, whereas all ranges of minimal wall thicknesses (e.g. first frame) and areas in which material accumulations are necessary (e.g. inner skin under the tail unit) can be distinguished by colour. NAFEMS Seminar: „Optimization in Structural Mechanics“ 7 April 27 - 28, 2005 Wiesbaden, Germany Fig. 10: Density distribution of the topology optimization based on a shell meshed design space This second step was performed based on the complete model, to allow a stiffness rearrangement e.g. from frame to frame. By conducting the optimization of the single component at this point, it can lead to changed loads and as a consequence to problems in other areas of the model 3.5 Following steps in the development process Another step in the development of the tail structure consists of the quantification of the solution. The topology optimization is first of all supposed to acquire qualitative material distributions, e.g. where are the best spots to place frames or connection elements. The question of ideal fin thicknesses can only be answered by using size- , shape optimization or a combination of both. The optimization model can be upgraded by regarding stress constrains, damage values of lifetime calculations or buckle safety factors [1]. A detailed description of the tail structure’s development process can be found in a separate publication [5]. 4 Conclusion The topology optimization with Altair OptiStruct was an important tool in the concept phase of the A400M airfreighter’s development process. Already in the early development stage, statements about the material distribution based on the given loads can be made in order to decide design questions based on fact. Further on tools around the optimization software take care that optimizations can be performed with complex FE-models and accordingly formulate the task deserved on demand, if it is about the definition of design variables, boundary conditions of the optimization and objective function. For this reason optimization tools are more and more integrated into the development process, in order to save cost-intensive and time consuming iteration loops. NAFEMS Seminar: „Optimization in Structural Mechanics“ 8 April 27 - 28, 2005 Wiesbaden, Germany 5 References [1] Altair OptiStruct, Users Manual v7.0, (2004), Altair Engineering inc., Troy MI. [2] Bendsoe, M.P.; Sigmund, O.; Topology Optimization – Theory, Methods and Applications, Springer Verlag, Heidelberg, 2003. [3] Schramm, U.; Structural Optimization – An efficient Tool in Automotive Design. ATZ – Automobiltechnische Zeitschrift, 100 (1998) Part 1: 456-462, Part 2: 566-572. (In German, English in ATZ worldwide). [4] Schramm, U.; Thomas, H. L.; Zou M.; Manufacturing Considerations and Structural Optimization for Automotive Components, (2002), Society of Automotive Engineers, Inc. [5] Schuhmacher, G.; Stettner, M.; Zotemantel, R.; O’Leary, O.; Wagner, M.; optimization assisted structural design of a new military transport aircraft, (2004), American Institute of Aeronautics and Astronautics. [6] Zou, M.; Shyy, Y.K.; Thomas, H. L.; Checkerboard and minimum member size control in topology optimization, (1999), Proceedings of the 3rd World Congress of Structural and Multidisciplinary Optimization, Buffalo, New York. NAFEMS Seminar: „Optimization in Structural Mechanics“ 9 April 27 - 28, 2005 Wiesbaden, Germany