Modern building design is complex and involves many different disciplines operating in a
fragmented manner. Appropriate computer-based decision support (DS) tools are sought
that can raise the level of integration of different activities at the conceptual stage, in order
to help create better designs solutions. This project investigates opportunities that exist for
using techniques based upon the Genetic Algorithm (GA) to support critical activities of
conceptual building design (CBD). Collective independent studies have shown that the
GA is a powerful optimisation and exploratory search technique with widespread
application. The GA is essentially very simple yet it offers robustness and domain
independence. The GA efficiently searches a domain to exploit highly suitable
information. It maintains multiple solutions to problems simultaneously and is well suited
to non-linear problems and those of a discontinuous nature found in engineering design.
The literature search first examines traditional approaches to supporting conceptual design.
Existing GA techniques and applications are discussed which include pioneering studies in
the field of detailed structural design. Broader GA studies are also reported which have
demonstrated possibilities for investigating geometrical, topological and member size
variation. The tasks and goals of conceptual design are studied. A rationale is introduced,
aimed at enabling the GA to be applied in a manner that provides the most effective
support to the designer. Numerical experiments with floor planning are presented. These
studies provide a basic foundation for a subsequent design support system (DSS) capable
of generating structural design concepts.
A hierarchical Structured GA (SGA) created by Dasgupta et al [1] is investigated to
support the generation of diverse structural design concepts. The SGA supports variation
in the size, shape and structural configuration of a building and in the choice of structural
frame type and floor system. The benefits and limitations of the SGA approach are
discussed. The creation of a prototype DSS system, abritrarily called Designer-Pro
(DPRO), is described. A detailed building design model is introduced which is required
for design development and appraisal. Simplifications, design rationale and generic
component modelling are mentioned. A cost-based single criteria optimisation problem
(SCOP) is created in which other constraints are represented as design parameters.
The thesis describes the importance of the object-oriented programming (OOP) paradigm
for creating a versatile design model and the need for complementary graphical user
interface (GUI) tools to provide human-computer interaction (HCI) capabilities for control
and intelligent design manipulation. Techniques that increase flexibility in the generation
and appraisal of concept are presented. Tools presented include a convergence plot of
design solutions that supports cursor-interrogation to reveal the details of individual
concepts. The graph permits study of design progression, or evolution of optimum design
solutions. A visualisation tool is also presented.
The DPRO system supports multiple operating modes, including single-design appraisal
and enumerative search (ES). Case study examples are provided which demonstrate the
applicability of the DPRO system to a range of different design scenarios. The DPRO
system performs well in all tests. A parametric study demonstrates the potential of the
system for DS. Limitations of the current approach and opportunities to broaden the study
form part of the scope for further work. Some suggestions for further study are made,
based upon newly-emerging techniques.
Date of Award | 2000 |
---|
Original language | English |
---|
Awarding Institution | |
---|
Optimisation and Decision Support during the Conceptual Stage of Building Design
Mathews, J. D. (Author). 2000
Student thesis: PhD