Date Approved

12-31-2002

Embargo Period

5-12-2016

Document Type

Thesis

Degree Name

M.S. in Engineering

Department

Electrical & Computer Engineering

College

Henry M. Rowan College of Engineering

Advisor

Mandayam, Shreekanth

Subject(s)

Nondestructive testing; Signal processing--Digital techniques; Pipelines--Maintenance and repair

Disciplines

Electrical and Computer Engineering

Abstract

Nondestructive evaluation (NDE) techniques offer cost-effective strategies for monitoring the integrity of a variety of civil infrastructure, such as natural gas and sewer pipelines, without the need to take the system off-line. However, interpretation of NDE signals in terms of the location, size, and shape of underlying flaws in the material being inspected is fraught with difficulty. Typically, variations in the testing signal due to operational parameters have created a significant challenge for defect characterization. This thesis proposes, develops, and validates a defect characterization algorithm that compensates for operational variables and maps the test signal to a visual defect profile. This algorithm takes a two-step approach:

  1. The raw NDT signal is processed via an invariance transformation feed forward artificial neural network that removes the effects of operational parameters and produces a signal containing defect related information only.
  2. A second feed forward artificial neural network processes the defect signature developed by the invariance transformation network and predicts defect profiles representing the location, size, and shape of material flaws.

The algorithm is validated with experimental data from two separate NDT sources, magnetic flux leakage (MFL) testing of metal gas pipeline specimens and ultrasonic testing (UT) of concrete wastewater pipeline specimens. A selection of three papers is provided describing the invariant defect characterization technique. The results obtained demonstrate that the invariance transformation technique can be used to accurately characterize the depth of defects in concrete or metal, irrespective of variations in the material properties of the test specimens. Recommendations for future research related to this technique are also provided.

Share

COinS