The sequential design optimization of building performance
Computational methods of design optimization offer great potential in solving the challenges posed by the
design of high-performing buildings. However, numerical optimization in building design industry is not
sufficiently exploited and has not yet emerged as part of the design process. This reveals the need to develop
optimization methods to support the identification of well performing designs while overcoming the long
computational time of performance evaluation due to the multi-dimensionality of real-world problems. In
response, this thesis aims to evaluate a sequential approach in the multi-objective design optimization of
building geometry, fabric, HVAC system and its controls.
An algorithm based on individual optimization steps conducted in sequence while concurrently considering
building design elements is developed to ensure a low computational burden and a whole-building holistic
approach. Three metrics are developed to evaluate the sequential design optimization: the performance of
the sequential search, the reliability (robustness) to variations of the same problem formulation with respect
to uncertain boundary conditions (weather and internal loads) and scale (four scales with increasing number
of solutions), and the efficiency of the search as compared performance with a NSGA-II algorithm. Six criteria
pertaining to three categories are applied to the three metrics for the evaluation: search effectiveness in
finding solutions with respect to the search and global optima, computational performance of the search with
respect to the computational load and computational savings compared to a full factorial search, and solution
optimality with respect to the objective and variable space. 24 different configurations of the search derived
from 4 initial starting points, 3 categories of grouping of the building design elements and 2 runs of the
algorithm are investigated.
The results are presented for the landscape of configurations of the sequential search and benchmarked
against the global set of optimal solutions derived from a full factorial search. The obtained results displayed
different magnitudes of performance, reliability, and efficiency depending on the configurations of the
search. Therefore, according to the purpose of the optimization, different configurations of the sequential
search could be preferred for different stages of the building design process. Among them, the iterative run
of the search in a 2-stage sequential process mapping the architectural and engineering disciplines (field
grouped) configured as (building geometry + fabric) and (HVAC system + controls) matched the results
obtained from a full factorial search in terms of effectiveness in finding solutions and solution optimality,
with a computational saving of function evaluations of 91.2% compared to the full factorial. Additionally, the
sequential search proved to be reliable to variations with respect to uncertain boundary conditions and scale
of the problem formulation. Finally, for an equal number of function evaluations, the sequential search
outperformed the NSGA-II in terms of effectiveness in finding solutions and solution optimality.
Funding
Engineering and Physical Sciences Research Council (EPSRC)
History
School
- Architecture, Building and Civil Engineering
Publisher
Loughborough UniversityRights holder
© Riccardo TalamiPublication date
2022Notes
A Doctoral Thesis. Submitted in partial fulfilment of the requirements for the award of the degree of Doctor of Philosophy of Loughborough University.Language
- en
Supervisor(s)
Jonathan Wright ; Bianca HowardQualification name
- PhD
Qualification level
- Doctoral
This submission includes a signed certificate in addition to the thesis file(s)
- I have submitted a signed certificate