Probabilistic models based on experimental observations using sparse bayes methodology

Kim, Minseo

Permalink

https://hdl.handle.net/2142/46861 Copy

Description

Title

Probabilistic models based on experimental observations using sparse bayes methodology

Author(s)

Kim, Minseo

Issue Date

2014-01-16T18:18:59Z

Director of Research (if dissertation) or Advisor (if thesis)

Song, Junho

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Sparse Bayes Methodology
• Bayesian Inference
• Probabilistic shear strength model

Abstract

The quality of people’s lives depends on safe and reliable infrastructure. However, there exist various types of uncertainties that may influence performance of structures, which could cause unexpected failures. Therefore, it is important to quantify the risk of such failures through systematic treatment of uncertainties and make a risk-informed decision. As an effort to predict uncertain performance of structural elements based on experimental observations, the principle of Bayesian inference has been often used. In this study, the recently proposed Sparse Bayes method is reviewed and tested by use of an experimental database of the shear strengths of reinforced concrete beams without stirrups. The performance of the Sparse Bayes method is demonstrated through comparison with existing methods such as least-square method and penalized least-square method. The Sparse Bayes method is further developed to identify a few representative points in the parameter space by grouping the relevant vectors identified by the method using the k-means clustering algorithm. The study confirms that the Sparse Bayes method has wide applicability, ability to achieve an optimal fitting, and efficiency. Therefore, the method can potentially provide a useful tool to develop powerful probabilistic models for various problems based on experimental observations.

Graduation Semester

2013-12

Permalink

http://hdl.handle.net/2142/46861

Copyright and License Information

Copyright 2013 Minseo Kim

Owning Collections