Session: Track 1-1B: Rail

Paper Number: 124287

124287 - Investigating the Use of B-Spline Signature Responses to Detect Internal Rail Defects 

The aim of this work is to explore the use of a proposed structural change detection method to detect internal rail defects. The unique nature of the railroad environment requires a damage detection technique to be robust enough to handle ambient noise and be easily implementable on long lengths of track. These particular requirements are the focus of this implementation.

The time-domain B-Spline transfer function is a signature time history response of a dynamic system. This signature response is defined as the response of the system to a cubic B‑Spline impulse excitation and may be either computed directly, or can be extracted from measurements of displacement, velocity, or acceleration time history responses to any external excitation. The signature response is referred to as the B-Spline Signature Response (BSR) in this work. The BSR is independent of loading, is a characteristic of the system, and captures the condition of the system at the time of acquisition without the need for identifying structural properties, e.g., stiffness, natural frequencies, etc. The BSR changes only when the condition of the system changes. The BSR’s invariance for an unchanged system is the basis of a proposed non-parametric data-driven change detection method, which utilizes cross-correlation to distinguish variance in two BSRs of the same system extracted at different points in time. System change is determined when the correlation of two BSRs drops below a defined threshold. The threshold value depends on the type and level of change being monitored.

The change detection methodology is demonstrated first in a deterministic manner through computer simulations with a numerical sensitivity study on an idealized dynamic system and through an experimental study on a steel plate supported by a braced frame. Through this study it is determined that: (i) the BSR method can detect change even with small changes in natural frequency, (ii) the method is more sensitive to detecting change when using displacement as the type of response, and (iii) the study reveals some patterns that could potentially be used for damage localization.

Next, the method’s sensitivity to uncertainty in response measurements is explored. A procedure to determine how sensitive the BSR method is to noise in a specific system’s responses is established. The procedure is demonstrated through computer simulations on noisy responses to harmonic excitation, first on a thin plate and subsequently on a rail model. In both cases, damage is considered as a transverse cut. Through this study, it was determined that the BSR has low sensitivity to noise in the responses to harmonic excitation of a rail.

In the current form, the BSR extraction process requires knowledge of both the response and excitation time histories. However, capturing the excitation time history is not a trivial task in any testing process. This paper presents the latest development that removes the dependency on the excitation time history. Through computer simulations of a rail, it is demonstrated that a decomposition of the excitation maintains the capacity of detecting system change.

Presenting Author: Brennan Gedney South Carolina University

Presenting Author Biography: TBD

Authors:

Brennan L. Gedney University of South Carolina
Reza Naseri University of South Carollina
Dimitris C. Rizos University of South Carolina