Event Title

A Neural Network for Depth Determination of Separations Between a Rubber Matrix and Reinforcing Steel Belts

Location

Snowbird, UT, USA

Start Date

1-1-1999 12:00 AM

Description

Neural networks have proven to provide a powerful technique to determine the crack size and orientation from ultrasonic backscattering data [1,2]. In this paper, we discuss a neural network for depth determination of separations between a rubber matrix and reinforcing steel belts. The reinforced rubber matrix is taken as a layered structure. The time-domain signals reflected from the layered structure containing separations at various depths have been simulated, and the simulated signals have been used as training data for the neural network. To evaluate a number of depth ranges, the neural network is comprised of several sub-networks, each of which covers a depth range of 0.61 mm with depth increments of 0.2 mm. A classifier that employs a cross-correlation algorithm is used to preprocess an input signal and to send the signal to the desired sub-network. The neural network has been tested on both simulated and measured signals.

Book Title

Review of Progress in Quantitative Nondestructive Evaluation

Volume

18A

Chapter

Chapter 3: Simulations, Signal Processing, Tomography, and Holography

Section

Classification Techniques

Pages

835-841

DOI

10.1007/978-1-4615-4791-4_107

Language

en

File Format

application/pdf

This document is currently not available here.

Share

COinS
 
Jan 1st, 12:00 AM

A Neural Network for Depth Determination of Separations Between a Rubber Matrix and Reinforcing Steel Belts

Snowbird, UT, USA

Neural networks have proven to provide a powerful technique to determine the crack size and orientation from ultrasonic backscattering data [1,2]. In this paper, we discuss a neural network for depth determination of separations between a rubber matrix and reinforcing steel belts. The reinforced rubber matrix is taken as a layered structure. The time-domain signals reflected from the layered structure containing separations at various depths have been simulated, and the simulated signals have been used as training data for the neural network. To evaluate a number of depth ranges, the neural network is comprised of several sub-networks, each of which covers a depth range of 0.61 mm with depth increments of 0.2 mm. A classifier that employs a cross-correlation algorithm is used to preprocess an input signal and to send the signal to the desired sub-network. The neural network has been tested on both simulated and measured signals.