Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.

org

Combining Parametric and Non-Parametric Structural


Optimization for Urban Air Mobility Conceptual
Development
Vishal Savane
(Dassault Systemes, India);

Rohan Keswani
(Dassault Systemes, France);

Torsten Moecker
(Dassault Systemes, Germany);

Rachel Fu
(Dassault Systemes, United States)

Abstract

With multiple startups and Original Equipment Manufacturers (OEMs)


competing to develop conceptual vehicles for Urban Air Mobility (UAM), new
tools and methods are necessary to streamline the design process to meet the
ambitious goals of this emerging industry. Determining appropriate lightweight
configurations for the airframe structure is a challenging task, as it requires
close collaboration between different disciplines, such as design, structural
optimization and flight mechanics. The focus of this paper is on proposing an
efficient workflow for conceptual structural sizing, which benefits from the
tight integration of parametric design, simulation and optimization capabilities.
To obtain the optimized lightweight design, parametric and non-parametric
optimization techniques are combined.

The workflow is entirely based on parametric design data, generated through a


combination of graphical visual scripting and interactive 3D modeling. This
method enables the engineer to create logic to parametrically build all required
components of the internal structure, including ribs, spars, frames, and
stringers. Fully associated with this parametric design model, a structural
model of the UAM vehicle is built, allowing for design space exploration.

In any given flight condition, the vehicle is subjected to aerodynamic loads,


rotor forces, gravity and inertia loads. Aerodynamic loads are determined using
a Computational Fluid Dynamics model, which is again fully associated with
design, while rotor forces are computed to ensure flight loads equilibrium.

Considering two exemplary load cases representing critical flight conditions,


parametric and non-parametric structural optimization techniques are then
combined, aiming at minimal weight for a targeted stress level. Based on

Presented at the NAFEMS World Congress 2021 1 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

results from a parametric design study, an optimized configuration of the


parametric UAM vehicle is determined. In addition, non-parametric sizing
techniques are applied, allowing for further reduction of mass by optimizing
the distribution of skin thicknesses and stiffener properties. In a final step,
structural requirement checks for buckling and strength are performed to
validate the optimized configuration.

1. Introduction

When it comes to the transportation of the future, a few technologies stand out
in particular: electric vehicles, autonomous vehicles, and vertical take-off and
landing vehicles (VTOL). The latter are frequently designed to be electrically
powered (eVTOL). A great deal of excitement and interest has gathered around
urban air mobility (UAM), which would involve using eVTOLs in urban
airspaces for both passenger and cargo transport, thus reducing ground traffic.

While the UAM industry is still getting off the ground, multiple startups and
OEMs are working to develop conceptual vehicles, creating an already-
competitive market. In order to stay ahead in this emerging industry, these
manufacturers must take advantage of new tools and methods to streamline the
design process, which is actually inherently multidisciplinary, due to the strong
interrelations between the involved design disciplines.

Closely linked to the use of electric propulsion for VTOL aircraft is the
onerous development of electric batteries as the energy source, with energy
density standing out as one of the major hurdles. Since sufficient battery
energy density may become the limiting factor for the success of the UAM
industry, one of the primary aims in the conceptual design phase is to
determine the most lightweight vehicle structure that meets requirements.

To achieve this goal, it is important to make best use of the existing


optimization technology for structural sizing. In particular, a tight integration
of parametric design, simulation and optimization technology is required to
facilitate fast design iterations and find an optimized structural configuration in
terms of weight. This paper proposes an efficient workflow for conceptual
structural sizing in order to address this challenge.

2. Structural Sizing Workflow

The workflow presented in this paper is applied to a tilting rotor conceptual


eVTOL design, which was previously developed as a demonstrator within
Dassault Systemes. This design, comprising two front counter-rotating open
rotors and two rear shrouded counter-rotating tilting rotors, is shown in
Figure 1.

Presented at the NAFEMS World Congress 2021 2 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Figure 1: eVTOL design used as baseline for structural sizing

Sizing of the external surfaces of an air vehicle based on aerodynamic


considerations is usually one of the first steps in conceptual design. In this
paper, however, we focus on the subsequent step of structural sizing.
Therefore, we assume that the geometry of the external surfaces is known.
Based on this input, the structural sizing workflow starts with developing a
parametric design model of the internal airframe structure and building a
structural model fully associated with this design model (see Section 3).
Associativity between the simulation as well as optimization technology and
the design environment is a key aspect which makes this workflow highly
efficient and flexible.

To illustrate the workflow, two exemplary load cases representing critical


flight conditions from a typical flight envelope are selected. Details on these
load cases are given in Section 4.

The overall objective of the structural sizing workflow is to find an optimized


lightweight design for the airframe structure. This is achieved by a combination
of parametric and non-parametric optimization techniques as described in
Section 5. We apply parametric optimization in order to determine an
optimized vehicle configuration in terms of number and location of internal
stiffeners, while non-parametric sizing techniques allow us to further reduce
the mass by optimizing the distribution of skin thicknesses and stiffener
properties. Finally, results are given in Section 6.

3. Parametric Airframe Model

In a first step, a parametric model of the internal airframe structure is built in


the design environment provided by the 3DEXPERIENCE® platform. The
external surfaces of the vehicle are assumed to be available as an input.
However, the parametric model is built in a generic way, which allows the
geometry of the external surfaces to be replaced by any update coming from

Presented at the NAFEMS World Congress 2021 3 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

the aerodynamics group in a later design phase. Replacing the geometry of the
external surfaces results in an automatic update of the internal airframe
geometry.

A combination of graphical visual scripting and interactive 3D modelling is


used to generate the parametric model of the internal airframe structure. The
image on the left hand side in Figure 2 shows the 3D modelling interface with
different internal airframe members such as ribs, front and rear spar, pylon
bulkheads, floor structure, front and rear bulkhead, centre wing box and
stringers. The image on the right shows the graphical modelling interface with
the graph view. Within the graph view, the parametric model is built by
dragging and dropping of existing components as well as by creating
connections between them. A full set of parameters is considered, e.g.
including the number of stringers and ribs or the location of front and rear
spars.

As the model is created for use in the conceptual design phase, no detailed
solid geometry is generated, but geometric representations by surfaces (e.g.
ribs, spars, bulkheads) and curves (e.g. stringers) are used.

Figure 2: 3D modelling (left) and graphical modelling (right) interface

Fully associated with this parametric design model, a structural finite element
model is built for use with the solver Abaqus/Standard®. Thanks to the
associativity, every change of a parameter in the design automatically updates
the structural model shown in Figure 3.

Presented at the NAFEMS World Congress 2021 4 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Figure 3: Finite element model of internal airframe structure (outer skin hidden)

The model consists of fully integrated conventional shell elements (type S4) as
well as linear beam elements (type B31). Shell elements are used for structural
members such as skin, ribs and spars while stringers are represented by beam
elements. Shared nodes are used to properly connect all parts of the structure.
The front and rear rotors are not modelled in detail, but are idealized with point
masses, which are connected to the structure through distributing coupling
constraints. Other nonstructural masses such as batteries and passengers are
considered using the same approach. Mass estimates are given in the next
section in Table 2.

It is assumed that the structure is made from composites. However, since a


detailed stacking sequence is usually not known in the conceptual design
phase, equivalent homogeneous material properties (see Table 1) are derived
from a quasi-isotropic layup of a standard carbon fibre reinforced polymer
(CFRP) material.

Table 1: Quasi-isotropic material data

Young’s modulus 50471 MPa

Poisson’s ratio 0.325

4. Flight Loads

For structural sizing, two exemplary load cases based on a typical flight
envelope are selected. The flight envelope defines the operational limits for an
air vehicle by specifying the range in terms of airspeed V and load factor n, in
which the vehicle can operate safely. Because no certification specifications
currently exist for eVTOL vehicles, we consider the certification specification

Presented at the NAFEMS World Congress 2021 5 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

CS-27 for small rotorcraft issued by the European Union Aviation Safety
Agency (EASA). According to section CS-27.337, it is required that “the
rotorcraft must be designed for a limit manoeuvring load factor ranging from a
positive limit of 3.5 to a negative limit of -1.0” [1]. The flight envelope used in
this study is shown in Figure 4. The cruise speed of 64.1 m/s (equivalent
airspeed) is determined under the assumption that the eVTOL vehicle operates
at a Mach number of 0.2 and at an altitude of 1000 m.

Figure 4: V-n diagram defining the flight envelope

As indicated in Figure 4, two load cases are selected from this flight envelope:

• Load case 1: Positive limit load at n=3.5 and V=64.1 m/s


• Load case 2: Negative limit load at n=-1.0 and V=64.1 m/s

For each load case, the following loads must be applied to the structure:

• Inertia load defined by mass distribution and acceleration of n⋅g


(g is the gravitational acceleration of 9.81 m/s2)
• Aerodynamic pressure load
• Rotor loads considered as concentrated forces

Inertia load

In order to properly account for the inertia load, it is required to add


nonstructural masses to the structural model as described in Section 3. Mass
estimates for all nonstructural components considered in this study are listed in
Table 2. Based on [2], the battery weight is approximated for a 4-passenger
vehicle.

Presented at the NAFEMS World Congress 2021 6 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Table 2: Mass estimates

Component Mass [kg]


Rear rotor [x2] 61 * 2
Front rotor [x2] 12 * 2
Battery 626
Passenger [x4] 77 * 4

The inertia load is applied using the inertia relief capability in


Abaqus/Standard® [3]. With this method, all externally applied forces
(aerodynamic pressure and rotor forces) are automatically balanced through a
load, which is proportional to the spatial mass distribution. The user does not
need to define a magnitude for the inertia load, but the magnitude is computed
internally by the solver to exactly ensure static equilibrium. To suppress any
unconstrained rigid body modes in the displacement solution, boundary
conditions are automatically applied at the centre of mass.

This approach is particularly useful in an optimization context: As the vehicle


mass changes in every optimization cycle, the magnitude of the inertia load
must be recomputed in every optimization cycle in order to avoid unbalanced
external loads. With inertia relief, this is achieved in an automatic manner.

Aerodynamic pressure

In the context of sizing the external vehicle shape, the aerodynamics team will
already have evaluated the vehicle’s aerodynamic performance through
Computational Fluid Dynamics (CFD) simulations. Similar to the structural
model described in Section 3, the CFD model shown in Figure 5 is also fully
associated with the design model. As both, the structural and the CFD model
are connected to the same design geometry, it is straightforward to map the
aerodynamic pressure distribution from the CFD mesh to the structural mesh
for reuse in the structural sizing task.

Figure 5: CFD model

Presented at the NAFEMS World Congress 2021 7 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

The aerodynamic performance of the vehicle is characterised by lift and drag


polars showing lift respectively drag as a function of angle of attack. From the
results of the CFD simulations, the resultant lift and drag forces given in
Figure 6 were derived.

Figure 6: Lift (left) and drag (right) polars (true airspeed: 67.3 m/s, altitude: 1000 m)

Rotor loads

Finally, the rotor loads are computed from equilibrium considerations. The lift
contributions from the wings and the rotors must equal the inertia loads
resulting from a load factor of n=3.5 (load case 1) respectively n=-1.0 (load
case 2).

� 𝐹𝐹𝑧𝑧 = 0 (1)

Since the rear rotors are designed as tilting rotors, they generate a thrust force,
which must balance the drag force from the aerodynamic pressure distribution.

� 𝐹𝐹𝑥𝑥 = 0 (2)

Moreover, in a trimmed configuration moment equilibrium about the pitch axis


must be respected.

� 𝑀𝑀𝐶𝐶𝐶𝐶,𝑦𝑦 = 0 (3)

From these equations and considering the polars in Figure 6, rotor forces to be
applied in the sizing load cases are determined. For both load cases, the flight
load equilibrium is summarized in Figure 7.

Presented at the NAFEMS World Congress 2021 8 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Figure 7: Flight loads for load case 1 (left) and load case 2 (right)

5. Structural Optimization

In the conceptual design phase, both, parametric optimization and non-


parametric sizing techniques are useful tools. Combining these methods allows
the engineer to make best use of the latest optimization technology and to
increase the potential for weight reduction. In this study, we illustrate how such
a combined optimization workflow can be built efficiently through a tight
integration of the optimization technology with the design environment.

Parametric Optimization

Since parametric optimization technology is available in the


3DEXPERIENCE® platform, it can easily be applied to the parametric airframe
model described in Section 3. In general, there are two approaches to
parametric optimization: either optimization algorithms are applied directly to
the parametric model, or they are applied to surrogate models which are built
from the results of a Design of Experiments (DOE) run. In this study, a
Pointer-2 optimization technique is applied directly to the parametric model.
The Pointer-2 technique automatically combines and controls a complementary
set of different optimization algorithms such as evolutionary optimization,
sequential quadratic programming, particle swarm and others. The design
space for the parametric optimization is defined by the following design
variables:

• Number of stringers (4…6)


• Number of ribs in centre wingbox (4…9)
• Number of ribs between pylon and fuselage (5…7)
• Number of ribs in outer wing (8…11)
• Thickness of centre wingbox skin (2 mm…4 mm)
• Number of pylon bulkheads (5…8)

Within each iteration of the optimization run, a non-parametric sizing task is


performed in order to further reduce the mass by optimizing the distribution of
outer skin thicknesses and stringer properties. Details on the non-parametric

Presented at the NAFEMS World Congress 2021 9 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

sizing set-up are given below. Figure 8 shows a diagram of this workflow
connecting parametric and non-parametric optimization.

Figure 8: Workflow combining parametric and non-parametric optimization

The objective of the parametric optimization is to minimize the total structural


mass 𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 subject to a stress constraint and a displacement constraint.
Additional stress constraints are accounted for in the non-parametric sizing part
of the workflow as explained below.

min(𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑙𝑙 )

𝜎𝜎𝑚𝑚𝑚𝑚𝑚𝑚,𝐶𝐶𝐶𝐶𝐶𝐶 ≤ 30 𝑀𝑀𝑀𝑀𝑀𝑀 (4)

𝑢𝑢𝑚𝑚𝑚𝑚𝑚𝑚 ≤ 100 𝑚𝑚𝑚𝑚

where 𝜎𝜎𝑚𝑚𝑚𝑚𝑚𝑚,𝐶𝐶𝐶𝐶𝐶𝐶 is the maximum stress in the centre wingbox and 𝑢𝑢𝑚𝑚𝑚𝑚𝑚𝑚 is the
maximum displacement of the structure.

Non-parametric Sizing

The non-parametric sizing task is performed with the sensitivity-based sizing


algorithm available in Tosca Structure [3]. Design sensitivities for this
algorithm are computed with the adjoint sensitivity method in
Abaqus/Standard® [3]. Using adjoint sensitivities is the preferred approach for
typical non-parametric optimization problems having a medium to large
number of design variables. Figure 9 summarizes the shell element thicknesses
and beam element cross section areas, which are considered as design variables
in this study, and also describes the corresponding thickness bounds.

Presented at the NAFEMS World Congress 2021 10 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Figure 9: Design space for non-parametric sizing

In this example, no clustering of elements in certain areas is defined, but a free


sizing optimization is performed. The optimization objective is to minimize the
structural mass subject to stress constraints:

min �� 𝑚𝑚𝑖𝑖 �

𝜎𝜎𝑚𝑚𝑚𝑚𝑚𝑚 ≤ 100 𝑀𝑀𝑀𝑀𝑀𝑀 for skin, spars and all beam elements (5)

𝜎𝜎𝑚𝑚𝑚𝑚𝑚𝑚 ≤ 30 𝑀𝑀𝑀𝑀𝑀𝑀 for centre wingbox skin

where 𝑚𝑚𝑖𝑖 is the mass of an element and 𝜎𝜎𝑚𝑚𝑚𝑚𝑚𝑚 is the maximum stress in a given
element set.

6. Optimization Results

The optimized configuration of the internal vehicle structure obtained from the
parametric optimization run is shown in Figure 10.

Figure 10: Optimized vehicle configuration

Presented at the NAFEMS World Congress 2021 11 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

For the final vehicle configuration, the optimization history of the non-
parametric sizing task is plotted in Figure 11. The plots indicate that the
objective function defined in Equation 5 approaches a minimum while the
stress constraints are met. The resultant thickness distribution in the outer skin
respectively the resultant distribution of beam element cross section areas is
shown in Figure 12. As expected, the maximum skin thickness occurs along the
primary load path from the pylons to the fuselage. For this example, combining
parametric and non-parametric optimization leads to a structural mass of
500 kg respectively a total vehicle mass of 1581 kg.

Figure 11: Optimization history for non-parametric sizing in terms of objective (left)
and constraints (right)

Figure 12: Thickness distribution in outer skin (left) and beam element cross section
areas in stringers and frames (right)

The final optimized configuration is then turned into a new design variant, and
additional structural requirement checks for buckling and strength are
performed to validate the configuration. In order to validate the requirement
that skin buckling may not occur below limit load, a linear eigenvalue buckling
analysis is performed. The lowest skin buckling loads are found at load factors
of 1.13 and 1.46 for the positive and negative limit load cases respectively. (A
load factor of 1.0 refers to limit load.) Figure 13 shows the corresponding skin
buckling modes.

Presented at the NAFEMS World Congress 2021 12 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Figure 13: Skin buckling modes for positive (left) and negative (right) limit load case

As discussed in Section 3, a simplified homogeneous material definition is


used in the optimization. For this reason, a more detailed strength check is
required to finally validate the optimized vehicle configuration. The
homogeneous material definition is replaced by a composite layup whose
stacking sequence and material properties [4] are given in Table 3. The
thickness distribution from the non-parametric sizing optimization is taken into
account by scaling the ply thicknesses accordingly.

Table 3: Composite layup and material properties [4] for final validation

Tensile fibre
Layup sequence (45, 90, -45, 0)s 1780 MPa
strength

Young’s modulus in Compressive fibre


133100 MPa 1405 MPa
fibre direction strength

Young’s modulus in Tensile matrix


9300 MPa 55 MPa
transverse direction strength

Compressive matrix
Poisson’s ratio 0.32 240 MPa
strength

In-plane shear
3745 MPa Shear strength 100 MPa
modulus

Considering this layup definition, a simple maximum stress failure criterion [3]
is evaluated for both load cases:

𝜎𝜎11 𝜎𝜎22 𝜎𝜎12


max � , , � �� < 1.0 (6)
𝑋𝑋 𝑌𝑌 𝑆𝑆

where 𝜎𝜎11 and 𝜎𝜎22 are the actual longitudinal stresses in fibre and transverse
direction, 𝜎𝜎12 is the actual shear stress and 𝑋𝑋, 𝑌𝑌 and 𝑆𝑆 are the fibre, matrix and
shear strengths of the material.

Presented at the NAFEMS World Congress 2021 13 25th-29th October 2021


© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Structural failure may not occur up to ultimate load, which is defined as


1.5 times limit load. This requirement is confirmed through the validation run,
as the evaluation of the failure criterion yields reserve factors of 2.5 and 9.1 for
the positive and negative limit load case respectively.

7. Conclusion

When it comes to the development of eVTOL vehicles, battery weight is one of


the limiting factors. Accordingly, there are significant weight requirements for
the vehicle structure to partially compensate for this limitation. To support the
engineer in reducing the vehicle weight by making best use of current
structural optimization technologies, this paper proposes an efficient workflow
for conceptual structural sizing. Tight integration of parametric design,
simulation and optimization technologies on the 3DEXPERIENCE® platform
is at the core of the workflow. Furthermore, parametric and non-parametric
optimization methods are combined, enabling the consideration of a wide range
of different design variables in a single optimization run. This includes
parametric design variables such as the number and location of structural parts,
as well as non-parametric design variables such as the element thicknesses.

To illustrate the workflow, it is applied to a tilting rotor conceptual eVTOL


design. Based on a parametric design model of the internal airframe structure
and considering two critical load cases, an optimized structural configuration
with a total weight of 1581 kg is found. Final validation runs confirm that this
configuration meets all requirements on composite strength and buckling.

8. References

[1] European Union Aviation Safety Agency (2020). Certification


Specifications and Acceptable Means of Compliance for Small Rotorcraft
CS-27, Amendment7: EASA.

[2] Bacchini, Alessandro; Cestino, Enrico (2019). Electric VTOL


Configurations Comparison: Aerospace 6, no. 3: 26.
https://doi.org/10.3390/aerospace6030026.

[3] Dassault Systèmes (2021). SIMULIA User Assistance 2021: Dassault


Systèmes.

[4] MIL-HDBK-17-2F (2002). Composite Materials Handbook, Vol. 2:


Polymer Matrix Composites Materials Properties: Department of Defense
Handbook.

Presented at the NAFEMS World Congress 2021 14 25th-29th October 2021

You might also like