LINGUISTIC DAMAGE ASSESSMENT USING FUZZY QUANTIFICATION THEORY (FATIGUE, WELDING)

ANDREW JOHN HINKLE, Purdue University

Abstract

Frequently in damage assessments, the available information includes linguistic descriptions of the damage and numerical results from tests and analyses. The precision of the numeric results and the precision of the linguistic descriptions may be inconsistent. However, the incorporation of both types of information is important for a meaningful assessment of structural damage. A linguistic damage assessment technique using fuzzy quantification theory is presented to incorporate both forms of information. The variables used in the assessment are described with linguistic terms which are represented by fuzzy membership functions. Fuzzy quantification theory is applied to a training sample to determine an assessment function. Using this function, the damage state can be estimated. The proposed method is demonstrated using experimental data on the effects of slag inclusions in butt welded joints. The training sample is randomly selected from a set of similar specimens and an analysis performed. The derived assessment function is then evaluated on the remaining specimens of the set. A criterion function based on the (a) fitness level, (b) accuracy of the assessment, and (c) width of assessed fuzzy set is proposed to determine the quality of the assessment function. The criterion function is used to examine the effects of (a) membership functions, (b) fitness level, (c) training sample, and (d) borderline variables on the assessment function. In addition, several conclusions are made and recommendations are presented.

Degree

Ph.D.

Subject Area

Civil engineering

COinS