Optimization to Body in White Design

APPLICATIONS OF OPTISTRUCT OPTIMIZATION TO BODY IN WHITE DESIGN

Carl Reed Jaguar Cars Cars Limited Body and trim CAE Engineering Engineering Centre Centre Coventry Coventr y CV3 4LF 4LF [email protected]

Abst Ab stract ract :

The application of topology and size and shape optimization for the design of efficient automotive components in terms of mass and stiffness has developed into a well practiced discipline. Subjecting a vehicle body structure to this type of development enables the analyst to optimise the size, shape and placement of load bearing members. By defining the most efficient topology, structural targets can be achieved with fewer design iterations, therefore, leading to reduced cycle times and lower development costs. The Altair OptiStruct software suite provides the analyst with an optimization environment which realises this method by providing a ‘right first time’ approach to body design. This paper presents an insight into the development of an efficient body structure subjected to various load cases. Based upon the topology of the load paths, which OptiStruct defines, the paper then details the development of the concept through to a detailed design.

Keywords:

1.0

Mass, Stiffness, Optimization, Topology, Optistruct, BIW, FEM

INTRODUCTION Increasing competition between automotive manufacturers has forced the development of new products to focus upon more efficient design. This leads to competitive product, increased sales and ult ultimately imately increased increased profits. profits. Each of the vehicl vehicle’s e’s systems systems contributes to its its overall efficiency. By assuming the body in white system behaves as a single component, topology optimization can be applied to ensure that material is distributed throughout the structure in an effici efficient ent manne mannerr to maximize maximize stiff stiffness ness and minimize inimize mass. The applicati application on of topology optimization to structural design provides an extremely fast route to ensuring the vehicle’s body structure meets or exceeds the relevant structural targets.

Copyright © Altair Engineering Ltd, 2002

5/1

2.0

OBJECTIVE The objective of this exercise is to develop a ‘right first time’ approach to the design of an automotive body structure. Once the method has been defined, a detailed concept, which matches or exceeds the structural targets, can be developed from preliminary styling surfaces with minimal design iterations. A design process has been developed, which enables the rapid development of BIW design at the concept stage of a vehicle lifecycle. The process developed below summarizes the key stages in the design cycle. (Figure 1). The details of the process are overviewed in the remainder of the paper.

Define External Boundaries (Styled A-Surface Concept)

Perform OPTISTRUCT topology optimization

Define space for occupant, power train, chassis

Define designable, nondesignable space

Create simplified beam and shell model

Perform OPTISTRUCT size and gauge optimization (NVH)

Create CAD model

Apply appropriate loads and weighting factors

Create detailed FEM

F i g u r e 1 : O p t i m i z at i o n D e s ig n P r o c e s s


5/2

Define section sizes (crash)

3.0

TOPOLOGY FEM AND DESIGNABLE SPACE DEFINITION When developing a new vehicle, the first data available to the engineering community is usually in the form of exterior surfaces released from Styling (Figure 2.0). At this stage of a vehicle’s development, information regarding its systems such as engine configuration, power train architecture, and chassis layout and occupant package is not always available. Hard points are not yet defined so assumptions based on the current vehicle architecture and or preferred use of components needs to be made in order to aid packaging structure. The more traditional approach to body development has largely been due to trial and error. By attempting to fit structure around a given package, components are tested in isolation and accepted or rejected due to their efficiency. Often the component does not work as well in isolation as it does in conjunction with others. When trying to find the optimum combination of load paths, the number of solutions required multiplies resulting in a time-consuming development matrix. Using OptiStruct topology optimization, the optimum material configuration of the BIW within the packaging space available can be determined. In order to achieve this, the available design space for the BIW must be generated as a solid FE model. At Jaguar a process has been developed to rapidly create this design space. Once the Styling surface is available, the first stage is to develop a solid mesh FEM which loads and boundary conditions can be applied to. By coating the released surfaces with a shell mesh, the total available volume for the structure can be created. From this, a sensible designable space model can now be constructed. (Figure 2.1)

Total available volume for optimization Figure 2.1: Initial Shell Mesh Enclo sing Th e Surfaces

Figure 2.0: Typical Scan Data Released from Styling

Next, the areas from which the body structure must be excluded need to be defined. These include the occupant cavity and space occupied by the engine power train and chassis. (Figure 2.2)


5/3

F i g u r e 2 .2 : E x c l u s i o n Zo n e s

D r i v e l i n e, E n g i n e , O c c u p a n t , W h e e l a n d B o o t Space

Obviously all that exists at this point is shell mesh which cannot be used for topology optimization. This now has to be converted into a solid mesh. The fastest method at jaguar to convert the closed volumes into a workable solid mesh is to use AKUSMOD. AKUSMOD is used extensively within the NVH analysis community to model fluid cavities within the occupant space for noise transfer function and sound pressure level predictions. In this application the meshing capability has been inverted (Figure 2.3). The solid mesh which is defined (the gaps between the exclusion zones and outer surfaces) becomes the allowable space for the body structure. (Figure 2.4). If the analyst does not have access to AKUSMOD, then the model could be generated with a solid meshing algorithm. However, this could prove to be time consuming as close attention would have to be paid to the surrounding shell mesh. AKUSMOD is good for this application because its meshing algorithm does not seed from the underlying shell mesh. Consequently, the analyst does not have to ensure that the volumes are fully enclosed. A mainly brick element mesh is generated, making it easier to perform any necessary hand editing.

Figur e 2.3: Model Set Up for AKUSMOD Meshing

Figure 2.4: AKUSMOD Mesh

Cavities excluded during AKUSMOD meshin .


5/4

To finish the solid model, some hand manipulation is necessary. By removing material around the doors, windshields, bonnet and boot space apertures, the basic OptiStruct topology is ready. (Figure 2.5)

F i u r e 2.5 : Fi n a li s ed A K U SM O D m es h

Remaining redundant material cut by hand

To complete the Optistruct model, the analyst is free to define areas of the mesh which are not to be modified by the optimization algorithm (i.e. essential structure). All vehicles require stiff longitudinal members for front and rear crashworthiness. These areas are therefore assigned to a non-designable Property ID. (Figure 2.6). When all of these have been defined, the Optistruct model is ready.

F i g u r e 2 .6 F i n a l i s e d T o p o l o g y D e f i n i t i o n

Designable space definition

Non-designable space definitions


5/5

4.0

SELECTION OF LOAD CASES FOR OPTIMIZATION An automotive body structure is subject to hundreds of different forces during every day of operation. However, many load cases are localised in nature and as such require detailed modelling in order to maximise their stiffness. When developing any structure it is always prudent to concentrate on the dominating load cases and their associated failure modes. As a consequence of selecting these, many of the local requirements will in turn be improved. The main load cases for a typical automotive structure can be categorised into two areas. 1. NVH type load cases - encompassing local and global body static stiffness. 2. Crashworthiness load cases - encompassing energy load path management.

4.1

LINEAR AND NON-LINEAR DOMAIN OPTIMIZATION Optimising topology for NVH type load cases is intuitive as the displacements involved are small. Since OptiStruct works within the linear domain it does not consider the failure mechanisms associated with crashworthiness such as large displacement, contact interactions and material behaviour beyond the elastic limit. In this instance, the analyst must recognise that the topology generated for any non-linear load cases must be regarded with engineering judgement. However, automotive crashworthiness targets dictate that the occupant cell acts largely as a rigid entity and therefore should not encounter any large displacement or material yield. This effectively renders it a linear system, and therefore all load distributing topology defined within the occupant cell is relevant.

4.2

LOAD CASES ANALYSED Each individual load case produces different topology with material distribution biased toward the load path between force and reaction point. By performing a separate optimization for each major load case and finding the common element density between them, the load cases can be weighted and optimised together. For example, all of the load cases contain the same mass (30% of the original), but the torsional stiffness element density topology comprises many of the load paths defined in the other load cases. This allows the load cases to be ranked according to their dominance. In this example, by applying a suitable weighting factor, all of the load cases can be optimised together. The following load cases weighting factors have been applied.


5/6

5.0

Ranking

Load Case

Attri bute

Weighting %

1

Torsion

NVH

45

2

Front crash

Crash

15

3

Side crash

Crash

10

4

Rear crash

Crash

10

5

Vertical bend

NVH

5

6

Rear end bend

NVH

5

7

Roof crush

Crash

5

8

Prop centre bearing

NVH

5

POST PROCESSING The results from the topology optimization indicate the placement of material in relation to the loads applied and their associated reaction points. When post processing the resultant structure, it is obvious which load cases have influenced the topology in certain areas of the vehicle. (Figure 3.0 & 3.1). Looking at specific areas such as the front end, definite triangulation around the front shock tower and sub frame can be observed.

Strong Cant rail and sill definition

Rear end bend brace

Triangulated shock tower F i g u r e 3 . 0: T o p o l o g y R e s u l t s


5/7

Looking at the structures rear end, material has been distributed from the cant rail down to the rear longitudinal members for rear end bend Further load paths from the front longitudinal members to the sills and tunnel have also been generated. Structural package tray and triangulated front sub frame have also been defined

Structural package tray

Triangulated front sub frame

Front crash load paths

F i g u r e 3 .1 : T o p o l o g y R e s u l t s

6.0

FURTHER MODEL DEVELOPMENT Now that the topology has been clearly defined, the next step is to create a simplified model to which real structural targets can be applied. Optistruct has been used to find the stiffest topology when subjected to a load by minimising the displacement response. To optimise to a real target, the displacement becomes the design goal. The fastest way to test the topology results is to covert the solid model into a simple beam and shell model. (Figure 4.0).

F i g u r e 4 .0 : B e a m a n d S h e l l M o d e l R e p r es e n t a t i o n o f T o p o l o g y R e s u l t s L e f t . Ri g h t B e a m and Shell Model


5/8

7.0

NVH OPTIMIZATION USING OPTISTRUCT Once the topology is fully defined, the structure needs to be optimised to a specific stiffness target. For this structure, the objective is to match the current best in class vehicle for torsional stiffness (BMW 7 350,000 NM/degree) whilst minimising mass. The design variables used for the optimization are based upon the dimensions of the beam and gauge of the shell elements. In isolation for the most efficient beam, its external dimension should be as large as possible whilst its wall thickness should be as thin as possible. As the constraint of minimum mass is defined, the optimum solution will not force all of the beams to the maximum dimension. In reality the physical constraints invalidate the mass vs. stiffness ratio. Each beam within the structure requires a certain stiffness. If the maximum dimension with minimum wall thickness is greater than that required, then the algorithm should reduce the external dimension and therefore reduce the section’s mass. This should cause a tapering effect across the structure according to the load path.

7.1

SECTION OPTIMIZATION SET UP The simplified model contains approximately 65 pairs of beams on opposite sides of the vehicle. By applying a different Property ID to each pair, individual design variables can be defined regarding the upper and lower control limits for the external diameter and wall thickness. Due to the small size of the model (approximately 1,400 elements), decomposition for each iteration takes only 5 seconds meaning converged solutions in under three minutes. By applying a vertical displacement target at the shock towers corresponding to the target torsional stiffness, the sections and gauges converge within 26 iterations. (Figure 5.0)


5/9

Initial condition

Iteration 5

Iteration 10

Restraints Loads

Iteration 20

Convergence – Iteration 26

Fig u r e 5 .0 : S e c t io n a n d G a u g e O p t i m i z a t i o n o f t h e B e a m a n d S h e l l M o d e l

Final mass = 100 kg Final stiffness = 350,000 NM/degr ee The optimization to a realistic target can only be defined in the linear domain, so for this optimization, only torsional stiffness has been optimised.

8.0

FRONT LONGITUDINAL OPTIMIZATION As previously stated, due to the non linear nature of crash mechanisms, linear topology optimization can only be used as a guide. To develop the crushable sections for front crash i.e. the front longitudinal members, the occupant structure is assumed to be rigid. Using LS DYNA, spring elements have been placed in position of the longitudinal members. (Figure 6.0). Their stiffness can be easily tuned to suit the target acceleration pulse.


5 / 10

F i g u r e 6 . 0: E a r l y F r o n t C r a s h A n a l y s i s t o T u n e C r u s h a b l e C o m p o n e n t s

When the stiffness of the longitudinal springs have been optimised to correlate to the target pulse. The size and shape of the longitudinal members can then be defined to match. (Figure 6.1)

Figure 6.1: Target Acceleration Pulse VS. Tuned B eam Model and Sub sequen t S e c t i o n R ec o m m e n d a t i o n

9.0

BIW GEOMETRY DEFINITION Once section sizes and panel thickness’ have been defined for both front impact and torsion, the generic beam model can then be converted into a CAD model. To develop the structure further now requires a more traditional detailed FEM based upon CAD geometry. Joint stiffness assembly techniques and panel gauge can now be investigated. (Figure 7.0)


5 / 11

Figu re 7.0 BIW Geometry Derived from Beam M o d e l

10.0

SEMI DETAILED FEM GENERATED FROM GEOMETRY By creating a semi detailed FEM, other linear and non-linear load cases can be applied. Joint definition and manufacturing techniques can also be selected. The concept retains many but not all of the proposed load paths. (Figure 8.0)

Crash Load paths as defined by OptiStruct

Triangulated Shock towers

Rear end bend braces Sections defined by beam model and crash work Parcel shelf definition by OptiStruct Figure 8.0: Semi detailed FE develop men t model


5 / 12

10.1

PRELIMINARY RESULTS When moving from the beam model to a more detailed FEM (containing joints and additional panels), the efficiency of the structure decreases. This is due to the beam model consisting of idealised joints. After Projecting the gauges from the beam model onto PSHELL cards and performing a shell thickness optimization, the semi detailed FEM achieves target stiffness (35,000NM/degree) at a mass of 165kg. This is approximately twice stiff as the structure it is designed to replace and less than half of its mass. The natural frequencies of the BIW structure are also approximately double its predecessor’s performance (Figure 8.1). However, these are likely to change as non structural BIW is added.

Kt = 35000NM/degree

Figu re 8.1: NVH Perfo rm ance – Lef t Mod al – Rig ht Tor sio n

In terms of crash performance, additional work is required to correlate the acceleration pulse to target. With the detailed FEM, the occupant structure is no longer assumed rigid so some additional gauge tuning is required which may increase the mass. However, in this example the failure mode of the energy absorbing longitudinal members has proved to be very stable and has provided a good starting point for further development. (Figure 8.2) Copyright © Altair Engineering Ltd, 2002

5 / 13

Excellent crush stability

Figure 8.2: Crash Performanc e

11.0

CONCLUSIONS The development of a ‘right first time’ CAE process for structural development has been demonstrated with the use of various optimization algorithms. Altair’s Optistruct optimization software has transformed the speed in which working CAE body concepts can be generated. Optistruct’s optimization algorithms have given the analyst the capability to distribute material across a structure in the most efficient way. This has enabled the development of a body structure topology which meets the structural targets in one design iteration. The BIW produced by this process is twice as stiff as the structure it is designed to replace and only half of its mass. The reduced number of iterations and subsequent model builds means development time can be cut to 8 weeks with only 3 analysts. Using more traditional methods takes a larger team of analysts several months and does not guarantee the optimum. Recent CAE methods of constructing detailed models identify failure mechanisms in more detail but the time taken to build them often means that results are late and hence recommendations are not acted upon. As a result CAE does not lead the design. For competitive vehicle performance and economy, one of the top priorities for vehicle manufacturers is to lower the mass of the product. In a typical vehicle, the body system makes up more than 25% of its gross weight. If this can be minimised, there is obviously a competitive advantage to be gained. By optimising the topology of the structure as described here, the material can be distributed in a more efficient manner thus giving equal or better structural performance at a lower mass.

12.0

ACKNOWLEDGEMENTS C Wright Jaguar Body CAE - Development of designable space method T Williams Jaguar Body CAE - Development of front longitudinal section


5 / 14

Optimization to Body in White Design

Recommend Documents