Damage detection of structural elements based on active sensing and machine learning approaches

  • Christabel Chepngeno Ng'etich Jomo Kenyatta University of Agriculture & Technology, Nairobi
  • James Kuria Kimotho Jomo Kenyatta University of Agriculture & Technology, Nairobi
  • John M. Kihiu Jomo Kenyatta University of Agriculture & Technology, Nairobi

Abstract

Monitoring the health of structural elements is very crucial in civil, mechanical and aerospace industries where accrual of small defects such as cracks may result in catastrophic failure. Nondestructive evaluation (NDE) techniques such as visual inspection, X-ray, ultrasound, eddy currents evaluation, magnetic particle testing, etc., are mostly used for damage detection and monitoring of structural components. Although these techniques are efficient in damage detection, they are time consuming, difficult, and expensive to apply regularly on structures. Structural health monitoring (SHM) techniques for damage detection by employing signals acquired directly from piezoelectric transducers attached on the structure have been developed to overcome this challenge. The main goal of SHM is geared towards the development of efficient approaches for processing and analyzing this data to obtain relevant information on the condition of the structure therefore providing real time monitoring. The main challenge of most SHM applications lies in establishing the right kind of signal that can be employed for effective detection of damage. As a contribution, this paper presents an approach for damage detection and damage classification in structural elements. This work incorporates signal processing, feature extraction, damage detection and application of machine learning algorithms for damage classification. A structural health monitoring test rig is set up consisting of a stainless steel cantilever bar instrumented with two Lead Zirconate Titanate (PZT) transducers for generation and detection of the ultrasonic guided waves for damage detection. Lamb waves and linear sine wave sweep are evaluated on their effectiveness in damage detection. The obtained sensor data is preprocessed and the extracted features used to train four machine learning algorithms with the obtained lamb wave sensor data achieving an average classification accuracy of 92.2% while the linear sine wave sweep sensor data achieved an average classification accuracy of 81.3%.
Published
Dec 4, 2019
How to Cite
NG'ETICH, Christabel Chepngeno; KIMOTHO, James Kuria; KIHIU, John M.. Damage detection of structural elements based on active sensing and machine learning approaches. JOURNAL OF SUSTAINABLE RESEARCH IN ENGINEERING, [S.l.], v. 5, n. 2, p. 62-77, dec. 2019. ISSN 2409-1243. Available at: <http://sri.jkuat.ac.ke/ojs/index.php/sri/article/view/756>. Date accessed: 31 mar. 2020.

Most read articles by the same author(s)

Obs.: This plugin requires at least one statistics/report plug-in to be enabled. If your statistics plugins provide more than one metric then please also select a main metric on the admin's site settings page and/or on the journal manager's settings pages.