Ijac20075403 Content
Ijac20075403 Content
net/publication/30870868
CITATIONS READS
118 5,191
6 authors, including:
All content following this page was uploaded by Dominik Holzer on 05 May 2017.
Parametric Design
and Structural
Optimisation
for Early Design
Exploration
Dominik Holzer, Richard Hough and
Mark Burry
626
03_Dominik 30/01/08 10:17 am Page 627
Parametric Design and Structural Optimisation for Early Design Exploration 627
03_Dominik 30/01/08 10:17 am Page 628
3. BACKGROUND RESEARCH
By investigating components of design problems, Lawson asserts: “since
design problems defy comprehensive description and offer an inexhaustible
number of solutions the design process cannot have a finite identifiable
Parametric Design and Structural Optimisation for Early Design Exploration 629
03_Dominik 30/01/08 10:17 am Page 630
4. RESEARCH METHODOLOGY
A live project has been selected to allow the authors with direct exposure
to an actual design problem with a high level of complexity.The project
(a stadium roof structure; Cox Group as architects, Arup as structural
engineers) was in an advanced stage of schematic design at the
commencement of the authors’ involvement, principally the lead author
who, as a qualified architect, was embedded within an engineering practice
(Arup, Sydney Australia).Working in this conceptual design stage facilitated
design input during design development of the project to test structural
performance for a variety of alterations to the geometry.The authors were
confronted with several unresolved design aspects which required
optimisation whilst bearing in mind structural stability, geometrical
Parametric Design and Structural Optimisation for Early Design Exploration 631
03_Dominik 30/01/08 10:17 am Page 632
Parametric Design and Structural Optimisation for Early Design Exploration 633
03_Dominik 30/01/08 10:17 am Page 634
strength checks for each member in a group. One constraint is active each
time while a series of checks is being carried out. Other methods would be
more rigorous but they cannot be applied for strength analysis as
constraints cannot be defined properly (they would work for displacement,
buckling or frequency analysis).The setup of any routine for the optimiser is
highly input-sensitive depending on a suitable initial choice of a set of
section sizes.The significance of this is explained below. It requires expert
input as one can otherwise easily get stuck in local minima.The main
challenge consists of optimising the section-sizes of a large array of
members individually under different loading combinations while aiming at a
global optimum for reducing the tonnage and maintaining structural stability.
Prior to commencement of the optimisation process, a limited set of
section sizes for specific sub-groups of the structure was chosen by the
engineers according to production constraints.The grouping occurred
according to design variables which dealt with effective length of the steel
members, their purpose in the structure, the architect’s requirements and
the aim to derive a nicely graded set of member-sizes.This resulted in the
definition of five groups of members:The groynes, the groyne-ties, the shells,
the front edge and the back edge. After initially selecting the smallest
section for each group of members to be optimised, the results were
compared to the requirements of the design codes applicable to the
project. If all the constraints were satisfied the optimisation was complete, if
not, the iterative process resizes those members which did not satisfy the
set criteria either up or down. Member size increments were limited to one
size increment per iteration. All results were communicated to an ‘MS
Access™’ database via an application programming interface that allowed
the engineers to read information in and out of the structural optimisation
software directly from their custom software. Results from the optimisation
process were obtained within a timeframe of approximately 30 minutes.This
assisted the research team in their effort to narrow the gap between
evaluating results and proposing changes for updating the parametric
geometry. Figure 6 displays a close-up of the corner-shell of the roof
structure after optimisation. Output from the ‘optimiser’ shows the varying
thickness of the grid-members for the roof as part of their cross-section
group.The steel members within the shell have the same outer diameter for
construction purposes, but the different strength-requirements need to be
picked up in varying wall-thicknesses of the hollow steel sections.
The structural engineers decided to first run tests to find the optimal
shape for the arches and then to subsequently focus on the curvature in the
individual shells. Once this was done, the structural engineers focused their
investigation on varying the curvature of the shells for the arch which
displayed the best results.
Observations of the results of the geometry variations led to the
proposal of a new (smallest) member size for one of the groups (shells)
Parametric Design and Structural Optimisation for Early Design Exploration 635
03_Dominik 30/01/08 10:17 am Page 636
which initially did not seem achievable.The structural engineers could see
that almost 90% of the members in the group were under-utilised. Detailed
information about the required diameter and length of steel members was
generated by the optimiser and could then be put out as an MS Excel™
spreadsheet and visualised in graphic tables as a by-product of the
optimisation process as seen in Figures 7 and 8.The information at hand
provided essential decision support for determining the direction in which
to alter the curvature sweeps of the stadium. Informed by the graphs,
coarse resolutions for the stadium roof geometry were derived initially and
then refined over time.
6. CRITICAL ANALYSIS
During the first three months of the development and testing of the
parametric model, the design team was logging all ongoing changes in the
stadium geometry. In this stage, a series of design parameters which were
assumed fixed, had to be altered and consequently the flexible model had
to be updated constantly.These changes were required due to aesthetic,
convergent, planning, or financial considerations and included inter alia a
revision of the main structural grid of the stadium, an alteration in the
position of the main roof supports and the variation of the extent of the
roof cantilever towards the pitch. Planning considerations were addressed
by investigating the ‘best fit’ of the parametric stadium model to the given
site-boundaries and the ‘high-ball-line’ being the minimum required field of
vision to a ball in play. In most cases, the changes could be accommodated
in the CATIA™ model, which led to a setup that was increasingly built on
dependencies. At the same time the complexity of these associative
dependencies increased, which had its effect on the hierarchical organisation
of design parameters within the CATIA™ file. As the main geometry of the
stadium was based on a rectangular grid and the arrangement of the roof-
support at the outside boundary of the stadium was based on a circular
array, some variations of the curvature led to complex intersections.
This required the introduction of transfer elements which had to be
accommodated in the parametric model retrospectively. For some variations
of the grid and shell curvature, the numerical definitions of the parameters
would not allow a possible solution to be rendered.This occurred when
design elements were over constrained by two or more design criteria.
The design team assessed that the intelligence derived from analysing
the decision making process which led to the alterations of the conceptual
geometry template, was more important than the geometrical ‘status’ of the
parametric model. As illustrated in Figure 9, one particular observation
made by the authors was that instead of an expected decrease of
indeterminate factors in the design, the number of variable design factors
increased during the three month design development. Because of this fact,
the necessity to scrutinise the design intent had become particularly evident,
which consequently assisted in developing the structural system in more
detail.This exemplifies that in the given case, ‘detailing’ in a parametric
context is neither a question of scale nor dependent on fixed parameters,
but rather depends on design logic and the correct parametric relations.
The fact that many numerical definitions of the stadium geometry were
unknown did not raise concerns as long as the parametric template could
accommodate them and meet the requirements of the design intent.
In addition to this, the structural engineers pointed out that the analysis
they required as decision support for understanding the structural behavior
did not have to be taken from the final ‘correct’ model, but it could be
generated from approximated parametric templates. As much as the
Parametric Design and Structural Optimisation for Early Design Exploration 637
03_Dominik 30/01/08 10:17 am Page 638
Parametric Design and Structural Optimisation for Early Design Exploration 639
03_Dominik 30/01/08 10:17 am Page 640
7. CONCLUSIONS
By linking parametric design to structural analysis and optimisation,
architects and structural engineers can explore design in the conceptual
design phase through informed geometry alterations.The setup of any such
flexible work environment requires a priori input from experts of both
professions to define suitability rules that guide the process towards a
specific performance goal.The rate of success depends on the precise
definition of quantifiable design variables across disciplines and the
awareness of the extent of variations being sought.The implementation of
this method and the type of parameters chosen are dependent on the
progress of the project according to the design stages. In the case
presented, it is neither aimed at finding an initial shape for the project, nor
for optimising the design for manufacturing and construction.The link
between parametric design and structural optimisation on the stadium
project has been applied in an advanced stage of conceptual design where
variations of a proposed design solution were sought by the team.
From an architect’s perspective, the immediate visualisation of structural
feedback provided by the structural engineers proved valuable to
understanding the effects of changes which might otherwise only be driven
by aesthetic considerations. In this context, immediacy and clarity of
information-display proved to be a decisive factor in facilitating shared
authorship.The opportunity of visualising and distributing results from
structural optimisation in close-to-real-time enabled the transdisciplinary
team to evaluate options and propose changes in a highly informed manner.
The more quickly results were communicated across a team, the better the
information flow and the collaborative capabilities.The graphic output of the
optimisation results gave a clear impression not only of the intelligence the
structural engineers are deriving from it, but also allowed the architects to
get insight into the working methodology of the engineers.
Tests on the stadium project have shown that if a project team relies
on automation routines within a project, members of the team require
access to information at any point in the design process for decision support
to guide the optimisation process and to propose alternative design
solutions.The automation routines run in the background as silent
partners whilst open access to the information helps avoid black-box
scenarios.The application of parametric variation was not only done to
benefit the structure, but it was actually necessary due to the complexity
of the structure at hand. Without the iterative experimentation, it would
have been impossible even for the experienced structural engineers to
understand the nature of the stress-distribution in the structure and to
consequently get a feeling for its behaviour.
The experience gained in the stadium project has served to underline
the need and definition for a more generalised software tool to act as a
data manager in an iterative design process, with a robust but flexible
8. FUTURE RESEARCH
The authors are currently extending their investigations based on the
findings described in this paper. Particular attention is being given to the
development of a framework for hosting the data manager as described
above and a corresponding data-storing mechanism to improve integration
and partly automate the processes for linking flexible geometry to building
performance.
Part of our ongoing research is the creation of a data manager for
closing the information loop between the geometry update and building-
performance optimisation.The idea of an intelligent feedback loop guided
through suitability rules assisted by the storage of geometrical and non-
geometrical information in a database will be central to the manager.
Acknowledgements
The authors would like to acknowledge the contribution of Yamin Tengono
from SIAL as well as Steve Downing, and Mark Arkinstall from Arup in the
conception of this paper and their contributions to the design team.
References
1. Robbin,T. Engineering a new Architecture, Yale University Press, New Haven and
London, 1996.
2. Coenders, J.L. and Wagemans, L.A.G. The Next Step in Modelling for Structural
Design: Structural Design Tools [Proceedings of the international symposium on
shell and spatial structures; theory, technique, valuation, maintenance] Madrid,
2005, 85-92.
3. Chaszar, A. Bridging the Gap with Collaborative Design Programs, Architectural
Design, vol 73, no 5,Wiley-Academy, 2003, 112-18.
4. Malkawi, A.M. Performance Simulation: Research and Tools, Performative
Architecture, Spon press New York, 2004.
5. Bedrick, J.I. and Rinella,T.O. Technology, process, improvement, and culture change,
AIA Report on integrated practice; 2006, Retrieved 11.01.2007 from URL:
www.aia.org/SiteObjects/files/8_Bedrick%20Rinella.pdf
6. Peters,T. F. Architectural and engineering design: two forms of technological thought on
the borderline between empiricism and science, Bridging the Gap: Rethinking the
Parametric Design and Structural Optimisation for Early Design Exploration 641
03_Dominik 30/01/08 10:17 am Page 642
Richard Hough
Arup
201 Kent St., Level 10, Sydney NSW 2000, Australia
[email protected]
Parametric Design and Structural Optimisation for Early Design Exploration 643
03_Dominik 30/01/08 10:17 am Page 644