Premium
Machine Learning User Preferences for Structural Design
Author(s) -
Thurston Deborah L.,
Sun Ruofei
Publication year - 1994
Publication title -
computer‐aided civil and infrastructure engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.773
H-Index - 82
eISSN - 1467-8667
pISSN - 1093-9687
DOI - 10.1111/j.1467-8667.1994.tb00372.x
Subject(s) - computer science , process (computing) , frame (networking) , engineering design process , machine learning , identification (biology) , multi objective optimization , optimization problem , artificial intelligence , industrial engineering , data mining , mathematical optimization , engineering , algorithm , mechanical engineering , telecommunications , botany , mathematics , biology , operating system
Abstract: The design process often proceeds through iterative stages of design configuration, analysis, evaluation, and redesign with the ultimate goal of optimization. Numerical methods for structural design optimization of only one attribute such as weight, strength, or cost are well known. However, these methods do not reflect the fact that designs are evaluated by the user in terms of their performance in several attributes. It has been extremely difficult to incorporate multiple attributes into design optimization algorithms because the acceptable tradeoffs between these attributes vary significantly between users. This paper presents a new method for learning user‐specific preferences and integrating them into the design evaluation, analysis, and optimization process in a meaningful way. The approach is a synthesis of formal decision theoretic methods with conventional design analysis techniques. The overall design objective is optimization of multiattribute utility from the viewpoint of the user. A user‐interactive computer‐aided Multiattribute Structural Design Evaluation and Optimization System (MSDEOS) is presented. It enables machine learning of the user's willingness to make tradeoffs between performance attributes. With this system, it is feasible to integrate site‐specific consideration of multiple attributes directly into computer aids for structural design optimization. Two examples are presented: seismic design, where tradeoffs are made between cost and damage index, and design of a three‐story steel frame structure, where attributes are cost and drift index. The system learns the preferences of different users and reflects those preferences through the identification of a different optimal solution for each user.