altair engineering - a holistic virtual design process applied … · 2019-04-17 · a holistic...

Copyright © Altair Engineering Ltd, 2002 13/1

A HOLISTIC VIRTUAL DESIGN PROCESS APPLIED TO AN A-PILLAR SUBJECTED TO ROOF CRUSH

S J Bates University of Wales Swansea

School of Engineering Singleton Park

Swansea SA2 8PP

[email protected]

J Randle Lea-Francis Ltd. 4 Bayton Road

Exhall Coventry CV7 9DJ

[email protected]

M Gambling Altair Engineering Ltd.

Vanguard Centre Sir William Lyons Road

Coventry CV4 7EZ

[email protected]

M R Jolly University of Birmingham

IRC Net Shape Manufacturing Lab.

Birmingham B15 2TT

[email protected]

Abstract:

The innovative design of a cast A-pillar for the Lea-Francis sports car, to meet Federal Motor Vehicle Safety Standard (FMVSS) 216, Roof Crush, is presented. The design was achieved with the use of a low-cost holistic virtual design process. This design process used the topology optimisation tool in Altair OptiStruct to generate an initial optimum topology for the A-pillar. Based on computer manufacturing simulations, experience, and the topology optimization results, an initial concept design was produced. The performance of this initial design was assessed using Implicit LS-DYNA. The model was then parameterised to enable final size and shape optimization. The resulting design is a lean structure, which is predicted to meet all its design targets. The manufacturing simulations have also given high confidence that the component can be successfully cast. A further benefit of applying this design process is that a greater understanding of the physical factors that drive the design has been obtained. The complete study was performed within two man-months, and with limited cost since all of the activities were entirely virtual. Keywords: OptiStruct, Topology Optimization, MagmaSoft, Implicit LS-DYNA, Size & Shape Optimization

1.0 INTRODUCTION

The Lea-Francis motor company was established in 1895, and launched its first car in 1903 (Figure 1a). Since the 1950’s Lea-Francis has been relatively inactive, lacking the investment required to compete with the larger manufacturers that were investing in cost-effective mass production designs. With the aim of introducing cutting-edge design, a new Lea-Francis sports car is currently being developed. A marketing prototype of the car has been built (Figure 1b) and the current styling has been further advanced (Figure 2a).

(a) First Lea-Francis 1903 (b) Latest Prototype Model

Figure 1: First and Current Lea-Francis Vehicles


Lea-Francis is very small company aiming at the low-volume niche market, and therefore cannot support the design overhead carried by the bigger car companies whilst achieving the desired return on investment. To survive, the company must avoid the costs associated with mass-market design, and also the inaccuracy and unreliability associated with niche market design. As a necessity therefore, the company must be an early adopter of new technology in order to meet these objectives. This paper demonstrates a new holistic design approach [1] applied to quickly and cost-effectively achieve the optimum design for the A-pillar of the latest vehicle. Virtual images of the latest Lea-Francis styling concept (Figure 2a) and the location of the A-pillar (Figure 2b) to be designed are presented.

(a) Virtual Image of Latest Prototype (b) A-Pillar Location

Figure 2: Latest Lea-Francis Prototype and A-pillar Location For convertible cars, the A-pillar design can be one of the biggest problem areas. The A-pillar is the confluence of many areas of the car, and in durability tests the A-pillar area is often the first to fail. Traditionally, A-pillar design uses steel, which leads to unwieldy designs that are often “patched-up” to meet requirements (Figure 3a). Also steel solutions require dedicated tooling, which is undesirable for a small company. A typical aluminium design (Figure 3b) consists of extruded members welded together. Research has demonstrated the availability of high strength aluminium alloys with yield stresses of around 400 MPa with 15% elongation [2]. There is little benefit in using high strength aluminium in a welded structure, because the welds will only be as strong as the basic aluminium used for the welding. However, a cast aluminium solution can take full advantage of the high strength alloys that are available. It was therefore decided that a casting of high strength aluminium would be used for the A-pillar of the Lea-Francis vehicle.

(a) Steel (b) Aluminium Extrusions

Figure 3: Typical A-pillar Designs


The objective for this study is to minimize the mass of the A-pillar casting, whilst: –

(i) Passing FMVSS 216 roof crush test requirements. (ii) Ensuring that the resonant frequency of the windscreen and A-pillar structure is

above 20 Hz in order to avoid the impression of “scuttle shake”. As both these objectives require high A-pillar stiffness, and the roof crush requirement is the most difficult to meet, optimising for the roof crush load case should automatically generate a design which passes the resonant frequency requirement. It was therefore decided to optimize the design for roof crush, and then to check that the resonant frequency requirements have also been met.

2.0 FMVSS ROOF CRUSH CERTIFICATION The roof crush test FMVSS 216 [3] defines the loading conditions that apply to A-pillar design. The loading device is an angled rigid platten that is lowered onto the A-pillar. The loading applied to this platten is increased up to a value of 1.5 times the unladen vehicle weight ( = × ×reaction unladen_ vehicleF 1.5 g m ).

This is achieved by moving the platten at 13 mm/s until the peak load is achieved, and during this time the displacement of the rigid platten must not exceed 127mm. The test device is oriented so that: – (a) “Its longitudinal axis is at a forward angle (in side-view) of 5 degrees below the horizontal,

and is parallel to the vertical plane through the vehicle’s longitudinal centreline.” (b) “ Its transverse axis is at an outboard angle, in the front view projection (Figure 4b), of 25

degrees below the horizontal.” [3]

(a) Results for Typical Failed and Successful Designs (b) Rigid Loading Platten Position and Orientation

Figure 4: Schematic Representation of FMVSS 216


Typical results are schematically represented in Figure 4a. The green line shows a successful design where the force requirement is met before the displacement limit is reached, whereas the red line reaches the displacement limit first, and therefore represents a failed design. In the case of a cast A-pillar, the most likely mode of structural failure would be for the casting to fracture before the FMVSS force limit has been reached. Consequently, a further constraint is put on the design: the component must have no plastic strain greater than 10% at the FMVSS peak force levels.

3.0 CONCEPT OPTIMIZATION The purpose of this stage of the process is to generate a concept design which makes the most structurally efficient use of the available package space. The topology optimisation tool in Altair OptiStruct takes the package space as its starting point, and generates an optimum material layout to react the given loading conditions within the specified constraints.

3.1 Generation of the OptiStruct Model

The package space (i.e. the designable space) available is shown in Figure 5 as the yellow region. The non-designable space consists of the attachments, the outer A-surface, the unloaded side (all red), and the body structure (grey). The assumption of making the unloaded side as non-designable was deemed satisfactory after performing a sensitivity study using OptiStruct [4]. The total mass of the package space material is 12.58 kg. The roof crush loading is represented in OptiStruct as a single static load case, applied as nodal forces in the direction of movement of the platten.

Figure 5: Package Space Identification (Designable (yellow), Non-Designable (red) and the Body Structure (grey) )


3.2 Concept Structural Architectures For the topology optimization set-up, the optimization objective is to maximize the structural stiffness of the A-pillar, subject to a design constraint which is specified as a target mass. The target mass is defined as the fraction of solid material to remain, e.g. a target mass of 0.3 means that 30% of the solid material is to remain. The A-pillar is to be designed as a casting. An important enhancement to OptiStruct is the addition of draw direction manufacturing constraints, where the material layout produced can be orientated to a particular direction [5]. Two draw directions are available; one accounting for castings where the mould is extracted in one direction, and one for split casting where the moulds are removed from either side of the casting. OptiStruct then allows only cavities that are open and aligned with the sliding direction of the die. The OptiStruct model was analysed to give the optimum solution for various combinations of target mass fractions and manufacturing casting constraints. These results are shown in Figure 6. The typical solutions show a definite load path along the bottom edge, from a loading point to the fixings. It should be noted that these results do not show the non-designable A-surface, which also carries load.

(a) Mass Target 30% (b) Mass Target 30%, Single Draw Direction

(c) Mass Target 30%, Single & Split Draw Direction

(d) Mass Target 20%, Single & Split Draw Direction

Figure 6: Topology Optimization Results for Various Combinations

of Target Mass and Draw Direction Constraints


3.3 Concept Sorting From the OptiStruct results (Figure 6), the next stage is to sketch interpretations of the design concepts suggested by the structure’s natural performance. Two sketches of possible designs were postulated (Figure 7); they are generally the same except that the design shown in Figure 7a has three ribs at the top of the section. Since this is a virtual design process, there is only a small overhead in running both these design concepts through the rest of the process. Using Altair OSSmooth, the geometry recovery phase of the interpreted design was initially performed on the concept presented in Figure 7a, resulting in the initial concept design shown in Figure 8.

(a) (b)

Figure 7: Sketches of the Underlying Features of the OptiStruct Results

(a) Concept Design (b) A-Pillar Cross-Section

Figure 8: Initial Concept Design

4.0 MANUFACTURE A fundamental issue to be addressed by any design is manufacturing feasibility: ‘Can we make it ?’. For the vast majority of manufacturing processes numerical technology is available to perform a virtual assessment on whether a design is feasible. Such assessments are an absolute necessity for any holistic design process. For the Lea-Francis A-pillar casting design, an initial review was undertaken using established casting design guidelines, followed by a virtual assessment of the casting process using the analysis code MagmaSoft [6].

4.1 Review Against Design Guidelines The initial OptiStruct concept design was reviewed against experience-based casting design guidelines. Typically, these guidelines relate to features within a design (eg. junctions, bosses


etc.) that are required to be casting-friendly. Consequently, a revised design was produced which involved two major alterations to the initial design;

(i) The uniformity of the cross section was improved. Section A-A (Figure 9), shows how the cross-section of the initial design is not well balanced, which could lead to shrinkage problems. A better solution is to improve the uniformity of the cross-section thickness as in the revised section A-A.

(ii) The ribs have been removed from the top of the A-pillar, and the other remaining ribs have been made thinner to reduce porosity at the junctions.

Figure 9: Initial Manufacturing Design Input

4.2 Solidification and Mould-Filling Simulations A solidification simulation using MagmaSoft [6] was performed for the initial design, Figure 10a, and the revised design, Figure 10b. This simulation assumes that the mould for the casting is instantly filled with material; this enables rapid assessment of problem areas exhibiting high porosity.

(a) Initial Design (b) Revised Design

Figure 10: Porosity Results From Instant Fill Solidification Simulation.

The simulation results for the initial design (Figure 10a) show regions of high porosity, especially around the ribs at the top of the A-pillar and in the walls. The revised design exhibits significantly lower levels of porosity. For example, the porosity levels in the upper wall of the A-pillar shown Figure 10(b) are the size of ‘pin-pricks’ and have negligible impact on the


structural performance, and porosity has been reduced at the junctions because of the thinner ribs. The remaining regions of porosity could easily be overcome with the addition of heat sink mechanisms or feed metal reservoirs in the mould.

Figure 11 shows a simulation of how the mould fills for one ingate position. The number of ingates and positions can be optimized at a later stage using Altair OptiCast [7], but at this stage this simulation gives a quick assessment of the design and may have indicated manufacturing complications with the design. The simulation shows that no ‘splashing’ of the molten metal has occurred and that the mould fills satisfactorily. A cold front has formed (blue area); this can be overcome by introducing more ingates.

(a) (b)

(c) (d)

Figure 11: Filling Simulation of the Revised Design

The conclusion from this stage of Manufacturing Feasibility assessment is that there is a high confidence that the revised design will be suitable for manufacture by casting.

5.0 CAE VERIFICATION AND OPTIMIZATION

5.1 Assessment of baseline design For the final phase of the design process a rigorous CAE verification was performed, in order to accurately predict the performance of the proposed design. This required the explicit modelling of a rigid platten as it deformed the A-pillar, and the resulting non-linear behaviour of the geometry and the material. At this stage the revised design (Figure 12) was assessed using the newly-implemented Implicit solver in LS-DYNA [8], which is well suited to the quasi-static


loading regime encountered in the roof crush test. The solution time for this particular problem was on average 15 minutes using a HP J-class machine.

Figure 12: Manufacturable Design

The mode of structural failure for the A-pillar cast component would be to fracture before the FMVSS displacement limit of 127 mm has been reached. Consequently, a plastic strain of 10% within the A-pillar was considered as the limit of the design and the yield stress was conservatively assumed as 250 MPa. The simulation results for the revised design predicted a reaction force of 28kN at the 10% plastic strain limit. The deflection at the 10% plastic strain limit was 90mm (Figure 13), and the deflection at the reaction force limit (17kN) was only 28.8mm. The A-pillar and windscreen assembly was also assessed for first resonant frequency. The first mode was found to be well above the 20Hz target. Therefore the design has passed with a considerable margin, which indicates that there is further scope for optimizing the design.

(a) Von Mises Stress Distribution (b) Reaction Force and Plastic Strains

Figure 13: Typical A-pillar Results at a Displacement of 90mm.

5.2 Size and Shape Optimization The topology optimization undertaken in Section 3 had already resulted in an optimum material layout. The next stage in the process is to use size and shape optimization to optimize individual parameters such as thicknesses and radii, thereby fine-tuning the performance of the design. Altair StudyWizard [9] was used to perform the size and shape optimization. The optimization objective was to minimise mass whilst satisfying the design constraints of plastic strain (= 10%), displacement (= 127mm) and reaction force (=17kN). The design variables


consisted of three shape variables; the thickness of the inner rail that runs up the length of the A-pillar, the thickness of a tapered section near to the fixing point with IP beam, and the outer wall thickness of the A-pillar (Figure 14).

(a) Inner Rail Thickness (Shape Variable 1)

4 mm 10 mm 15 mm

(b) Tapered Section Thickness (Shape Variable 2)

4 mm 18 mm 32 mm

(c) Outer Wall Thickness (Shape Variable 3)

4 mm 6 mm 8 mm

Figure 14: Shape Variable Definition (Left to Right : Minimum, Baseline, Maximum)

For both the baseline and optimized models, the design objective, constraints and shape variables are presented (Tables 2 and 3). The geometric differences between the manufacturable and the optimized design are presented in Figure 15.

Design Variable

(DV)

DV Minimum

(mm)

DV Maximum

(mm)

Baseline Model (mm)

Optimized Model (mm)

Shape variable

1 4.0 15.0 12.0 4.0

Shape variable 2 4.0 32.0 27.0 12.0

Shape variable 3 4.0 8.0 6.0 4.0

Table 2: Shape Variable Comparison (Baseline and Optimized)


Objective & Constraints Baseline Optimized

Mass (kg) 4.50 3.88

Plastic strain (%) 5.0 8.5

Displacement (mm) 28.2 60.0

Reaction Force (kN) 17 17

Table 3: Objective and Constraint Comparison (Baseline and Optimized)

(a) Baseline (b) Optimized

Figure 15: Cross-Section of Baseline and Optimised Design

(Red = shape variable 1; Green = shape variable 2; Blue = shape variable 3) The optimization study also provides sensitivity information relating to how design variables affect the objective and constraints (Figure 16). Design variables 1 and 3 have the greatest effect on the displacement (a) and mass (c). Variable 2 has little effect on the displacement and mass responses, but contributes over 90% to the plastic strain response. These results provide useful information for the designer. Typically, an increased car weight (eg. different engine) will produce an increase in the FMVSS 216 reaction load limit. To cope with this increase, the designer needs only change design variable 2, if the plastic strain is undesirable or change design variables 1 and/or 3 if the mass/displacement is undesirable.

(a) Displacement (b) Maximum plastic strain (c) Mass

Figure 16: Sensitivity of the System Parameters with Shape Variables


6.0 CONCLUSIONS An innovative A-Pillar design has been produced, which meets the FMVSS 216 roof crush requirements. Manufacturing considerations have been built into the design, and a high confidence level that the design can be successfully cast has been achieved. The design contains innovation in both concept and process, and has been optimized for mass to produce a lean structure. Through this process, a greater understanding of the physical factors that drive the design has been obtained. The complete study was performed within two man-months, and with limited cost and risk, since the design at this stage is entirely virtual.

7.0 REFERENCES

[1] 'Spotlight Innovation, Not Just Evaluation', R D Jones and D L Simon, Concept to Reality Magazine, Altair Engineering Limited, Spring 2002. [2] “Castings”, J.Campbell, Butterworth Heinemann, Oxford 1991 [3] 'FMVSS 216 section 571.216', Federal Motor Vehicle Safety Standards. [4] 'Altair OptiStruct Version 5.0', Altair Engineering Limited, 2001 [5] 'HyperMesh Version 5.1', Altair Engineering Limited, 2002. [6] 'MAGMASOFT’ MAGMA GmbH, Germany. [email protected] [7] 'Get the Most from your Casting Process', L E Smiley, Concept to Reality Magazine, Altair Engineering Limited, Spring 2002. [8] 'LS-DYNA : Non-linear Dynamic Analysis of Structures in Three Dimensions', LSTC, Version 960, 2002. [9] 'StudyWizard Version 5.0', Altair Engineering Limited, 2001.

altair engineering - a holistic virtual design process applied … · 2019-04-17 · a holistic...

Documents