Date of Graduation
1999
Document Type
Thesis
Degree Type
MS
Committee Chair
Powsiri Klinkhachorn
Abstract
Non-destructive damage detection using vibration techniques provides a valuable tool in identification of damage and monitoring of structural integrity in structures. Earlier research had shown that use of strain energy mode shapes allows identifying and locating damage. A highly qualified expert is required in order to analyze acquired mode shapes and recognize damage locations.
This research investigates several aspects of building an automated damage detection system utilizing the method of strain energy mode shapes. Such a system would replace a human expert at the most critical part of damage detection process. The study had been conducted using a FE model of simple beam and a limited number of experimentally acquired data.
Generation of reliable baseline mode shapes derived from a group of specimens with unknown damage state was studied. Several methods were suggested to perform this task. Besides, it was shown that the same methods can be applied to reduction of random measurement noise.
Several numeric metrics were introduced in order to study sensitivity of strain energy method. It was shown that even in conditions without measurement noise, insensitive to damage zones exist. A group of methods based on modification of amplitude spectrum of strain energy mode shapes was developed for improvement of damage detection sensitivity. It also proved possible to avoid usage of baseline mode shapes by applying modified versions of those methods.
It was shown that several mode shapes should be used for damage peak search and that correct identification of damage peaks is impossible if this procedure is performed using just magnitude information (thresholding). A human – like approach was studied and encoded into a rule-based fuzzy expert system. The expert system allowed reliable damage detection both for baseline and non-baseline methods.
The developed methods were tested on an extensive set of FE model experiments and a limited number of experimentally acquired data. For every tested damage scenario, method (baseline or non-baseline) and data origin (FE model, impact testing or laser vibrometer), damage location was successfully detected.
Recommended Citation
Sazonov, Eduard S., "A case study for building an automated damage detection system." (1999). Graduate Theses, Dissertations, and Problem Reports (ETD). 10441.
https://researchrepository.wvu.edu/etd/10441