Dissertation/Thesis Abstract

Parametric Attribution and Decomposition Methodologies for Product Model-Based Thermal Simulation using Multidisciplinary Design Optimization (MDO) Environments
by Welle, Benjamin Ross, Ph.D., Stanford University, 2012, 170; 3520399
Abstract (Summary)

Research has demonstrated that multidisciplinary design optimization (MDO) processes that automate the workflow from an object-oriented computer-aided design (CAD) model to performance simulation engines can compress design cycle time, increase design knowledge, and yield substantive product quality and performance gains. However, the process of developing and implementing a fully-automated CAD-centric performance simulation workflow for use in an MDO environment, in particular for thermal simulation, is methodologically and technically very complex with little guidance from existing methods in research or in practice. The up-front costs in developing such a process are high, both financially and in terms of time and human resources. If the MDO process developer is not successful in creating a process that supports flexible problem formulation, the result will likely be a rigid MDO process that lacks accuracy, scalability, and cost-effectiveness.

Flexible problem formulation for CAD-centric MDO may take many forms. There are two types of flexible problem formulation with major gaps in research that are the focus of this research. First, there are currently no CAD-centric attribution methods for MDO that support parametric attribution, or flexibility in how building attributes are associated with geometry to generate an attributed model for the desired optimization configuration. This functionality is critically important in order to ensure that the optimization is configured appropriately for the design problem it’s being used to evaluate. A failure to do so results either poor accuracy or poor cost-effectiveness in design space exploration, or both. Second, no automated methods exist to efficiently reduce the size of an attributed model that must be analyzed to accurately represent whole-building performance. Such a method would significantly reduce simulation time requirements for an MDO process, in particular for large buildings.

This research fills these gaps with the development of a CAD-Centric Attribution Methodology for Multidisciplinary Optimization Environments (CAMMOE) and a methodology for automated product model decomposition and recomposition for climate-based daylighting simulation called the BIM-Centric Daylight Profiler for Simulation (BDP4SIM). Both of these contributions to knowledge enable flexible MDO problem formulation and are implemented within an MDO workflow for automated BIM-based multidisciplinary thermal simulation called ThermalOpt. ThermalOpt integrates and automates the workflow from a parametric BIM (Digital Project) to an energy simulation engine (EnergyPlus) and a daylighting simulation engine (Radiance) using a pre-processor based on the open data model Industry Foundation Classes (IFC).

Validation of the CAMMOE and BDP4SIM methodologies with several test cases and two industry case studies provides evidence for their scalability, generality and power, as well the improved speed, accuracy, scalability, and cost-effectiveness they enable for MDO in AEC. The validation of CAMMOE with two diverse industry case studies demonstrated that without the availability of this methodology, the design teams for both projects would have only been able to either evaluate <1% of the design space they wanted to explore, resulting in extremely poor design space exploration accuracy, or would have been required to evaluate orders of magnitude more design options than were necessary, resulting in poor cost-effectiveness. The best performing designs from trade studies run over several distributed and parallel computing platforms using CAMMOE resulted in decreases in annual operating costs for energy of 16.5% and 24% relative to the design team’s final design. The validation of BDP4SIM with two test cases and an industry case study resulting in the methodology correctly identifying the minimum number of unique space for simulation with accuracies (DSerr) of 0%, 0.3%, and −0.1% (within the target ±2%) and reductions in simulation times of 69%, 76%, and 60%, respectively.

Indexing (document details)
Advisor: Fischer, Martin, Haymaker, John, Bazjanac, Vladimir
Commitee:
School: Stanford University
School Location: United States -- California
Source: DAI-B 73/10(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Civil engineering, Mechanical engineering, Energy
Keywords: Building information modeling, Daylighting simulation, Decomposition, Energy simulation, Multidisciplinary design optimization, Parametric attribution, Product models, Thermal simulation
Publication Number: 3520399
ISBN: 9781267493699
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest