Using MDO to Support Sequential Design Decisions Using MDO to Support Sequential Design Decisions Michael Lepech (PI) Martin Fischer (co-PI) Forest Flager John Basbagill CIFE TAC 2013 1 Copyright 2013 Using MDO to Support Sequential Design Decisions Case Study: Multi-family Residential CIFE TAC 2013 2 Copyright 2013 Using MDO to Support Sequential Design Decisions Conceptual Design Problem SCOPE Housing buildings: substructure ● shell ● interiors ● services OBJECTIVES (1) Minimize life-cycle cost (2) Minimize carbon footprint VARIABLES (1) Number of buildings: 3 - 4 a (2) Number of stories: 5 - 8 (3) Building footprint: H-shape c d f e b (4) Building orientation: 0-360° CONSTRAINTS (1) Gross Floor Area (GFA): 15,000m2 (2) Building width (a, b and f): >15m (3) Distance to site perimeter: >20m (4) Distance between building: >20m CIFE TAC 2013 3 Copyright 2013 Using MDO to Support Sequential Design Decisions Design Exploration Today: Is this sufficient? DESIGN SPACE • 1.46E11 possible design alternatives CURRENT INDUSTRY PRACTICE • 3 design alternatives evaluated • 162 man hours required CIFE TAC 2013 4 Copyright 2013 Using MDO to Support Sequential Design Decisions Introduction to MDO “Multidisciplinary Design Optimization (MDO) involves the formalization of design iteration and coordination to leverage computer processing power to systematically search the design space.” - MDO Technical Committee, AIAA (1991) GEOMETRY MODEL DProfiler CONSTRUCTION COST OPERATIONAL COST EMBODIED IMPACT OPERATIONAL IMPACT DProfiler eQUEST, CostLab SimaPro eQUEST, SimaPro OPTIMIZER Genetic Algorithm CIFE TAC 2013 5 Copyright 2013 Using MDO to Support Sequential Design Decisions MDO Case Study Results Life-Cycle Cost (USD, Millions) 225 KEY + Baseline X Selected Designs 210 Infeasible Solution 3 Buildings, 5 Stories 3 Buildings, 6 Stories + 195 3 Buildings, 7 Stories 3 Buildings, 8 Stories 4 Buildings, 5 Stories 180 4 Buildings, 6 Stories 4 Buildings, 7 Stories 4 Buildings, 8 Stories 165 275 280 285 290 295 Carbon Footprint (met kilotons CO2e) CIFE TAC 2013 6 Copyright 2013 Using MDO to Support Sequential Design Decisions MDO Case Study Results Life-Cycle Cost (USD, Millions) 225 KEY + Baseline X Selected Designs 210 Infeasible Solution 3 Buildings, 5 Stories 3 Buildings, 6 Stories + 195 3 Buildings, 7 Stories 3 Buildings, 8 Stories 4 Buildings, 5 Stories 2 180 1 4 Buildings, 6 Stories 4 Buildings, 7 Stories 4 Buildings, 8 Stories 3 165 275 280 285 290 295 Carbon Footprint (met kilotons CO2e) CIFE TAC 2013 7 Copyright 2013 Using MDO to Support Sequential Design Decisions CIFE MDO Research PAST KBZ NATIONAL STADIUM Scope: ANALYSIS LAYER Scale: 1:290.3 Steel roof structure Savings: USD $4M (19%) Support: 2009 CIFE Seed, Autodesk z y x EDUCATIONAL BUILDING Scope: Steel superstructure Savings: USD $570K (21%) Support: Skanska GOVERNMENT OFFICE BUILDING MULTI-FAMILY RESIDENTIAL Scope: Scope: Structural system Savings: USD $1.7M (7%) Savings: USD $4M (19%) 2012 CIFE Seed, Beck Support: 2009 CIFE Seed, Autodesk Support: Structural system CURRENT • MDO class under development • Ongoing MDO research CIFE TAC 2013 8 Copyright 2013 Using MDO to Support Sequential Design Decisions Conventional vs. MDO Design Process CONVENTIONAL - system type - system topology - building massing - orientation CONCEPT DEVELOPMENT MDO new information - component sizing DETAIL new information - building massing - orientation - system type - system topology - component sizing CIFE TAC 2013 9 Copyright 2013 Using MDO to Support Sequential Design Decisions Research Question How can automated design exploration be leveraged to support sequential design decisions? CIFE TAC 2013 10 Copyright 2013 Using MDO to Support Sequential Design Decisions Research Goals Develop a decision support methodology to provide stakeholders with clear and quantitative information about: Q1: What decisions are most important? A: Analyze relative influence that different decisions have on performance criteria of interest. Influence (%) 10 5 0 Glazing Thickness Cladding Material WWR Wall R-value Height Design Decision CIFE TAC 2013 11 Copyright 2013 Using MDO to Support Sequential Design Decisions Research Goals Q2: What is the impact of the decision on product performance? A: Analyze how a given decision affects the distribution of possible solutions. Probability of Occurrence (%) KEY No decisions made 18 1. Six floors 14 1. Six floors 2. Orientation = 323° 10 1. Six floors 2. Orientation = 323° 3. Steel cladding 6 2 150k 250k 350k 450k Impact (kg CO2e) CIFE TAC 2013 12 Copyright 2013 Using MDO to Support Sequential Design Decisions Research Goals 4 8 30 30 30 50 30 360 2.97E5 2.34E8 3 5 10 10 0 10 0 0 2.75E5 1.67E8 Number of Buildings Number of Floors Length a (m) Length b (m) Length c (m) Length d (m) Length e (m) Building Orientation (deg) Carbon Footprint (kg CO2e) Life Cycle Cost ($) Q3: What are the synergies between design decisions? A: Analyze the relationship between coupled decisions. Preference Shading: Worst CIFE TAC 2013 Best 13 Copyright 2013 Using MDO to Support Sequential Design Decisions Research Methods 1) Formalize design problem and process 2) Develop design space analysis methods - Design of Experiments (DoE) - Optimization 3) Develop data visualization methods 4) Implement decision support methodology - Massively parallel (cloud) computing environment CIFE TAC 2013 14 Copyright 2013 Using MDO to Support Sequential Design Decisions Research Validation 1) Parallel design charrettes / case studies - Subjects: students and practitioners - Groups: conventional (control) / decision support (intervention) 2) Sample metrics - solution quality / feasibility - design cycle duration - total design cycles - use of data visualization methods CIFE TAC 2013 15 Copyright 2013 Using MDO to Support Sequential Design Decisions Schedule Task 1: DEVELOP DESIGN SPACE ANALYSIS METHODS Timeline Summer 2013 2: DEVELOP DATA VISUALIZATION METHODS Fall 2013 3: IMPLEMENT DECISION SUPPORT METHOD Fall 2013 4: DESIGN CHARETTES 5. JOURNAL PUBLICATIONS CIFE TAC 2013 Winter 2014 Spring 2014 16 Copyright 2013 Using MDO to Support Sequential Design Decisions Research Impact CIFE 2015 Goals Schedule Performance dramatically reduced design cycle time Cost Conformance / Sustainability many alternatives evaluated quickly better performing designs identified early quantitative assessment of quality risks IMPROVED TIME TO MARKET COST CONFORMING, SUSTAINABLE DESIGNS CIFE TAC 2013 17 Copyright 2013 Using MDO to Support Sequential Design Decisions Opportunities for Industry Participation Owners / Designers / Contractors partner on case studies participate in design charrettes access to software / training USE OF METHODS / TECHNOLOGY TO SUPPORT BUSINESS OBJECTIVES Software Providers partner to incorporate your software into MDO environment Learn from new methods developed and user testing IDENTIFY NEW TECHNOLOGY / METHODS FOR PRODUCT DEVELOPMENT CIFE TAC 2013 18 Copyright 2013 Using MDO to Support Sequential Design Decisions Questions? CIFE TAC 2013 19 Copyright 2013 Using MDO to Support Sequential Design Decisions Budget Item Description Cost Post‐Doctorate Researcher Flager, Forest $26,179 Graduate Student Research Assistant TBD $45,626 TOTAL AMOUNT REQUESTED CIFE TAC 2013 $71,805 20 Copyright 2013