Sie sind auf Seite 1von 8

PROGRESS IN TOPOLOGY OPTIMIZATION WITH MANUFACTURING CONSTRAINTS M. Zhou, R. Fleury, Y.K. Shyy, H. Thomas, J.M.

Brennan Altair Engineering, Inc. 2445 MacCabe Way, Suite 100, Irvine CA92614 zhou@altair.com
Abstract Topology optimization has been shown to be an extremely powerful tool in generating efficient design concepts in the early stage of a design process. Unfortunately, very often designs suggested by topology optimization turn out to be infeasible for certain manufacturing process. At such occasions, it is often very difficult, if not impossible, to transform a design proposal to a manufacturable design. In this paper, design requirements for casting and extrusion production are addressed for topology optimization. Introduction Topology optimization has seen rapid development during the last decade, both in research and industrial applications [1-20]. The power of this technology lies in its early impact in a design process. It has been shown that topology optimization can help to create highly efficient design concepts. This very often leads to much more significant design improvement compared to sizing and shape optimization that can be only applied to a structure with given layout. The fast development of commercial software contributed to a rapid penetration of this technology in the industry. Most commercial FEA software have added certain capabilities of topology optimization. In this development, Altair OptiStruct [23], a commercial product that specializes in design optimization, has emerged to be a front runner in this direction. Several capabilities are unique to this product. It uses a general setting of a multiple constrained optimization problem, in which topology, shape and sizing optimization can be handled simultaneously [18]. Minimum member size control provides a means to control the degree of simplicity of the solution of topology optimization, which could be interpreted as a general purpose manufacturing constraint [17]. At the conceptual design stage, manufacturing feasibility is one of the key requirements. For example, casting is a very popular manufacturing means for mass production of machine parts. To be feasible for casting, cavities in the structure has to be open and lined up with the sliding direction of the dies. Unfortunately, topology optimization very often creates cavities that would prohibit drawing the dies. In many cases, it is impossible to transfer such a design concept into one that is feasible for casting production without losing its constructive merits. In a recent paper by Zhou, Shyy and Thomas [19], a general mathematical approach has been proposed for casting manufacturing constraints. In this paper, practical applications of the proposed method are shown. This capability has been implemented in the commercial software Altair OptiStruct. It is applicable to any 3D finite element model with arbitrary mesh. This capability is recently released with OptiStruct 5.1 [23]. Furthermore, requirements for extrusion, which is also a popular manufacturing procedure for mass production, are addressed in this paper. For extruded parts, the cross-section has to be constant along the path of extrusion. Topology optimization problem The general optimization problem can be stated mathematically as follows
Minimize f ( ) Subject to g j ( ) 0 i U gj 1, 0, j = 1,..., M i = 1,..., N

(1)

Where f ( ) represents the objective function, g j ( ) and g j represent the j-th constraint response and its upper bound, respectively. M is the total number of constraints; i is the normalized material density of the i-th element. Note that the problem in (1) is a relaxation formulation of the topology problem, where the density
U

Copyright 2002 by M. Zhou, R. Fleury, Y.K. Shyy, H. Thomas, J.M. Brennan Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

should only take the value 0 or 1. To enforce the design to be close to a 0/1 solution, a penalty is introduced to reduce the efficiency of elements with intermediate densities. For the SIMP approach [6][7] the penalization is achieved by the following power law formulation:
K ( )= i i p i Ki

Thus the optimization problem shown in (1) should be modified in the following manner:

Min f ( ) U S.t. g m ( ) i g m j 0, ... n (0 m = 1,...,M (4) 1) , k = 1,...,K k

(2)

where Ki and Ki represent the penalized and the real stiffness matrix at full density of the i-th element respectively, and p is the penalization factor that is bigger than 1. Typically p takes value between 2 and 4. In general, the optimization problem in Eq.(1) involves a very large number of design variables. However, the number of active constraints is usually small, if local constraints such as stress constraints are excluded. Because of this characteristic, the problem can be solved very efficiently by the dual method of nonlinear programming [21], [22]). Casting Manufacturing Constraints The mathematical formulation for casting constraints has been presented in [19], which is summarized below. Considering the finite element model shown in Fig.1, where the mesh is perfectly lined up in the sliding direction, the following constraints can sufficiently prevent the creation of cavities and boundaries that prevent the slide of the die:
i j i, ... n

where K represents the number of sets of elements that are lined up in the sliding direction of the die. It can be shown that the additional single variable linear constraints can be handled efficiently in the dual optimizer similar to side constraints [21]. A more typical scenario in casting is to have two dies pulling apart along the sliding direction. For this case, there are two sequences of variables along each line, separated at the die splitting point. Therefore, the above formulation does cover this situation when the die splitting surface is known. In OptiStruct, a method has been developed to optimize the die splitting surface under this two-die scenario [23]. Also handling of irregular mesh, e.g. tetra mesh, has been implemented. The details for these techniques are not presented in this paper. Different draw directions can be applied to subdomains of the packaging space for topology optimization. This can be useful in achieving so-called casting patches that are sometimes applied to complicated parts. Extrusion Manufacturing Constraints For the model shown in Fig.1, the requirement for extrusion in the sliding direction is
i = j = ... = n

(3)

where

j ,..., n represent densities of elements that

are along the same line in the sliding direction, as shown in Fig.1.

(5)

Which represents a simple variable linking. Thus the optimization problem (1) is modified as follows:

Min f ( ) U S.t. g m ( ) g m 0, (0 m = 1,...,M (4)

= = ... = 1) , k = 1,...,K n i j k where K represents the number of sets of elements that are lined up along the path of the extrusion. This problem actually represents a simplification of the original problem (1) in that the number of independent design variables is significantly reduced. This method is implemented in OptiStruct for the forthcoming release.
Figure.1 Illustrative finite element model

Numerical Examples Four numerical examples are presented to illustrate the impact of manufacturing constraints discussed in this paper. Example 1 Beam under Torsion with Casting manufacturing constraints: A coarse mesh of the beam has been used in reference [19] for study during the research phase. Here a finer finite element mesh, shown in Fig.2, is used and the structure is optimized using the released code OptiStruct 5.1 [23]. The beam has a dimension of 4x4x16, and 2048 hexagonal elements are used. One end of the beam is fixed and a pair of twisting forces is applied on the free end. The compliance is minimized subject to a volume fraction constraint of 0.3. Without casting manufacturing constraints we obtained a tube like structure shown in Fig.3, which is reached after 12 iterations. This design is indeed the optimal topology under torsion. However, if the beam has to be produced with a single die drawing upwards (in the Z direction), the tube solution does not give any idea for a casting feasible design. The solution under the given casting constraints is shown in Fig.4, which is achieved after 73 iterations. Periodical

Fig.4 Optimized topology with casting constraints X cells are formed to carry the twisting forces to the supported end. Because of the existence of semi-dense elements, a comparison of the performance of the final designs should be done with new FE models recovered from the final results. This could be done by post processing a result using OSSmooth embedded in Altair HyperMesh [24] to generate iso-surfaces for a given density threshold. During this process, OSSmooth also generates a tetra mesh for reanalysis. However, it is very difficult to find appropriate density thresholds for two designs that would result in the same material volume for comparison. As an alternative, analyzing the final result without penalty for intermediate density can be a reasonable approximation. This implies from Eq.(2) that the stiffness of an semi-dense element is considered to be linearly proportional to its material density. Eliminating penalty the compliance for the designs with and without casting constraints are 1.37688 and 2.44428, respectively. This means that the casting feasible design has a torsional stiffness of 56% of the design without casting constraints. Despite of the significant loss in stiffness compared to the tube design, the casting feasible design does appear to be an optimal concept under the given manufacturing constraints. The concept of using X cells to transform the twisting forces eliminates bending at the system level. Example 2 Engine Bracket with Casting Manufacturing Constraints: An aluminum engine bracket of a car has been optimized using earlier release of Altair OptiStruct and a reduction of weight from 950 g to 739 g has been achieved in the finalized product [20][17]. Since this is an aluminum cast part, it represents a good example for investigating the

Fig.2 FE model of the beam under torsion

Fig.3 Optimized topology without casting constraints

influence of the casting constraints during topology optimization. The finite element model of the design domain is shown in Fig. 5, in which 9046 elements are used and the design domain is shown in blue color. Six load cases were considered, which reflect the following driving and service status: 1) start; 2) backwards; 3) into a pothole; 4) out of a pothole; 5) loads from an attaching part and 6) loads during engine transport. The sum of compliance of all load cases is minimized for a given volume fraction of 0.3. The result without casting manufacturing constraints is shown in Fig. 6, which is reached after 25 iterations. An minimum member size constraint of dmin=15 mm is used, which is close to the default minimum member size of 16.50 mm determined by the average mesh size when casting constraints are applied. With a casting draw direction upwards (in the

Z direction), the result of a single die is shown in Fig. 7 and the result of two dies splitting is shown in Fig. 8, which needed 26 and 25 iterations, respectively. Reanalyzing the final designs without penalty for intermediate density provides a reasonable baseline for performance comparison. The compliance of the design without casting constraints is 52.7582. The compliance of the designs with casting constraints are 55.0611 for the single die option and 51.4573 for the two die splitting option. Comparing the reciprocals of the combined compliance, the values for the casting designs with a single die option and two die splitting option are 95.8% and 102.5%, respectively, of the value of the design without casting constraints. Note that the

Fig.5 Finite element model of the engine bracket

Fig.6 Topology of the engine bracket without casting constraints

Fig.7 Topology of the engine bracket with casting constraints single die

relatively small differences in compliance may fall into the range of errors due to allowance of semi-dense elements in the model. Nevertheless, the results indicate that for this design problem, the additional casting manufacturing constraints did not reduce the performance of the optimized design. Clearly, the design proposals with casting manufacturing constraints are much easier to interpret.

Example 3 Rail with Extrusion Requirements: A curved beam is considered to be a rail over which a moving load is applied (Fig. 9). 59151 elements, mostly hexagons, are used in the FE model. Both ends of the beam are simply supported. The moving load is simulated as a point load applied over the length of the rail as five independent load cases. The rail should be manufactured through extrusion. The objective is to minimize the sum of the compliance under all load cases. The material volume fraction is constrained

Fig.9 FE model of a rail under moving loads

Fig.8 Topology of the engine bracket with casting constraints two dies splitting

Fig.10 Topology of the constraints

rail without extrusion

at 0.3. The optimized topologies without and with extrusion constraints are shown in Fig.10 and Fig.11, respectively. 27 and 35 iterations are needed, respectively. Reanalyzing the final designs without penalty for intermediate density, the compliance for the designs without and with extrusion constraints are 29.9396 and 37.4377, respectively. This implies a 20% loss in performance due to extrusion constraints. Note that the extrusion design represents a clean proposal that requires little refinement. On the other hand, the design obtained without manufacturing constraints may require significant modification that could cause efficiency loss in performance.

Fig.12 FE model of sitting area of a folding chair minimum member size of 0.2 in is used. For the run with a draw direction perpendicular to the sitting surface, a default minimum member size of 0.46 in is active, which is calculated based on the average mesh size by OptiStruct [23]. The optimized topology without draw direction constraints is achieved after 38 iterations, which is shown in Fig.13. It can be seen that the structure forms a sandwich box underneath a portion of the area under pressure load. Some rather arbitrary rods are formed to connect the top and the bottom shells. It is clear that this design proposal does not give many clues about rib layout. Fig.11 Topology of the rail with extrusion constraints Example 4 Layout of Stiffening Ribs of a Folding Chair: The design problem is to find the optimal layout of stiffening ribs underneath the sitting surface of a folding chair. For this purpose, the shell structure is filled with solids in the FE model shown in Fig.12. A single load case simulating the sitting pressure load is considered. The dimension of the model is about 10x8.5x1.2 in. The compliance is minimized subject to a volume fraction constraint of 0.1 for the solid domain. For the run without draw direction constraints, a The optimized topology with draw direction constraints is reached after 52 iterations, which is shown in Fig.14. The result provides us with a clear design of the layout of stiffening panels. For a comparison of their structural performance, both final designs are analyzed without penalization of intermediate density. The compliance for the structures without and with draw direction constraints are 1.00449 and 0.94585, respectively. This implies, to our surprise, a 6% higher performance for the design with draw direction constraints. In general, it is true that additional constraints should reduce the feasible design region and hence result in a

higher objective function at the optimum. However, the topology optimization problem in (1) is a highly none convex problem due to the penalty applied. The existence of local optima could prevent the iterative process to achieve a better design, although the formulation without draw direction constraints does not exclude the design shown in Fig.14. The same phenomenon has also been found with the additional constraint on minimum member size [17].

Fig.14 Topology of the folding chair with draw direction constraints Altair OptiStruct 5.1 that has been released in June 2002 [23]. In addition, the formulation and solution of extrusion constraints are introduced. This capability has also been implemented in OptiStruct and will be available in the forthcoming release. Four applications demonstrate that solutions under manufacturing constraints are usually non-intuitive and could not be easily derived from topology designs obtained without taking manufacturing constraints into consideration. It is worth mentioning that both casting and extrusion manufacturing constraints could be used for conceptual design study of structures that do not need to be manufactured using the corresponding procedures. Those requirements could be regarded as specific geometric constraints and can be used for any design that desires such characteristics. Example 4 showed that the casting constraints could be used to locate the layout of rib patterns of a shell structure. If it is desired to have ribs going through the entire depth of a solid domain, extrusion constraints can be used to determine the layout of the stiffening panels. It is strongly recommended that when manufacturing constraints are desired, the user should run a baseline comparison with a formulation without those additional constraints. This could help to assess the relative efficiency loss due to the manufacturing constraints. This provides a trade-off study between cost and structural performance in the decision process for the

Fig.13 Topology of the folding chair without draw direction constraints Concluding Remarks It has been shown that manufacturing constraints are important aspects that require attention during the conceptual design stage. Since topology optimization serves as a tool for generating design concepts, incorporation of manufacturing constraints has significant impact in shortening the distance between concept and reality. In this paper, extension and applications of casting manufacturing constraints are presented. This capability has been implemented in the commercial software

right manufacturing procedure for the structure under consideration. There are more manufacturing constraints that should be addressed during topology optimization. One that has been frequently requested by users of OptiStruct is a maximum member size control. For example, this requirement is important for casting or extrusion parts, where the thermal dissipation requirements impose restrictions on member thickness. It appears that the more manufacturing constraints are considered for topology optimization, the closer to the final product reality the solution can be. A virtual prototyping directly out of topology optimization represents the ultimate challenge. References: Bendse, M. 1995: Optimization of structural topology, shape and material. Springer-Verlag, Berlin 2. Rozvany, G.I.N.; Bendse, M.; Kirsch, U. 1995: Layout optimization of structures. Appl. Mech. Rev., 48, 41-119 3. Sigmund, O.; Petersson, J. 1998: Numerical instabilities in topology optimization: A survey on procedures dealing with checkerboards, meshdependencies and local minima. Struct. Optim., 16, 68-75 4. Thomas, H.L., Zhou, M., Schramm, U. 2002: Issues of commercial optimization software development. Struct. Multidisc. Optim., 23, 97-110 5. Bendse, M.; Kikuchi, N. 1988: Generating optimal topologies in optimal design using a homogenization method. Comp. Meth. Appl. Mech. Engrg. 71, 197-224 6. Bendse, M. 1989: Optimum shape design as a material distribution problem. Strrut. Optim., 1, 193-202 7. Zhou, M.; Rozvany, G.I.N. 1991: The COC algorithm, Part II: topological, geometry and generalized shape optimization. Comp. Meth. Appl. Mech. Engrg., 89, 197-224 8. Sigmund, O. 1994: design of material structures using topology optimization. Ph.D. Thesis, Department of Solid Mechanics, Technical University of Denmark 9. Diaz, A.R.: Sigmund, O. 1995: Checkerboard patterns in layout optimization. Struct. Optim. 10, 40-45 10. Haber, R.B.; Bendse, M.; Jog, C. 1996: A new approach to variable-topology shape design using a constraint on the perimeter. Struct. Optim. 11, 1-12 11. Ambrosio, L.; Buttazzo, G. 1993: An optimal design problem with perimeter penalization. Calc. Var. 1, 55-69 1.

12. Jog,C.S.; Haber, R.B. 1996: Stability of finite element model for distributed-parameter optimization and topology design. Comp. Meth. Appl. Mech. Engrg. 130, 203-226 13. Petersson, J.; Sigmund, O. 1998: Slope constrained topology optimization. Int. J. Num. Meth. Engrg. 41, 1417-1434 14. Duysinx, P. 1997: Layout optimization: a mathematical programming approach. DCAMM Report No. 540, Technical University of Denmark 15. Olhoff, N.; Rnholt, E.; Scheel, J. 1998: Topology optimization of plate-like structures using 3-D elasticity and optimum 3-D microstructures. Proc. 7th AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and Optimization, St. Louis, Missouri, 1853-1863 16. Bendse, M.; Sigmund, O. 1999: Material interpolation schemes in topology optimization. Arch. Applied. Mech., 69, 635-654 17. Zhou, M.; Shyy, Y.K.; Thomas, H.L. 2001: Checkerboard and Minimum Member Size Control in Topology Optimization. Struct. Multidisc. Optim., 21, 152-158 18. Zhou, M.; Pagadipti, N.; Thomas, H.L. and Shyy, Y.K. 2000: An integrated approach to topology, sizing and shape optimization. Proc. 8th AIAA/NASA/USAF/ISSMO Symposium on Multidisciplinary Analysis and Optimization, Long Beach, CA, September, 2000 19. Zhou, M.; Shyy, Y.K.; Thomas, H.L. 2001: Topology optimization with manufacturing constraints. Proc. 4th World Congress of Structural and Multidisciplinary Optimization, Dalian, China, June, 2001 20. Schramm, U. 1998: Structural optimization an effective tool in automobile design, Part II. (in German) Atomobiltechnische Zeitschrift, 100. Jahrgang/Nr.7/8, 566-572 21. Fluery, C. 1989: CONLIN: an efficient dual optimizer based on convex approximation concepts. Struct. Optim., 1, 81-89 22. Svanberg, K. 1987: Method of moving asymptotes a new method for structural optimization. Int. J. Num. Meth. Engrg., 24, 359-373 23. Altair OptiStruct 2002: Users manual. Altair Engineering, Inc., Troy, MI 24. Altair HyperMesh2002: Users manual. Altair Engineering, Inc., Troy, MI

Das könnte Ihnen auch gefallen