Premium
Stochastic identification of defects under sensor uncertainties
Author(s) -
Furukawa Tomonari,
Lim Shen Hin,
Michopoulos John G.
Publication year - 2011
Publication title -
international journal for numerical methods in engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.421
H-Index - 168
eISSN - 1097-0207
pISSN - 0029-5981
DOI - 10.1002/nme.3283
Subject(s) - identification (biology) , computer science , kalman filter , parametric statistics , data mining , entropy (arrow of time) , probability density function , measure (data warehouse) , algorithm , mathematics , artificial intelligence , statistics , botany , physics , quantum mechanics , biology
SUMMARY This paper presents a new methodology for identifying defects under the presence of both sensor and defect uncertainties. This methodology introduces a representation of the beliefs of both the locations of defects and the sensors each by a probability density function and updates them using the extended Kalman filter. Because the beliefs are recursively maintained while the sensor is moving and the associated observation data are updated, the proposed approach considers not only the current observation data but also the prior knowledge, the past observation data and beliefs, which include both sensor and defect uncertainties. The concept of differential entropy has been introduced and is utilized as a performance measure to evaluate the result of defect identification and handle the identification of multiple defects. The verification and evaluation of the proposed methodology performance were conducted via parametric numerical studies. The results have shown the successful identification of defects with reduced uncertainty when the number of measurements increases, even under the presence of large sensor uncertainties. Furthermore, the proposed methodology was applied to the more realistic problem of identifying multiple defects located on a specimen and has demonstrated its applicability to practical defect identification problems. Copyright © 2011 John Wiley & Sons, Ltd.