Neural networks for pattern recognition
Neural networks for pattern recognition
Neural Networks: A Comprehensive Foundation
Neural Networks: A Comprehensive Foundation
Multi-resolution model fusion for corrosion classification
Second international workshop on Intelligent systems design and application
| Hi-index | 0.00 |
The paper presents an integrated methodology for detection and classification of corrosion that can be used to effectively characterize corrosion in critical areas such as lap joints for the aging aircraft. The approach involves a wavelet analysis based corrosion feature extraction method from the nondestructive inspection (NDI) data, and the extracted wavelet features are used by a neural net classification model for detection and classification of corrosion. From past experience, it has been seen that any given sensor is often insufficient for corrosion detection as each sensor has its own limitations. The experimental results from this project clearly validate that doing data fusion of multiple sensor signals improves the probability of detection and better assessment of corrosion damage. As part of our approach, we present a data standardization process that helps to bring data from different sensors to a common standard or scale, so that they can be compared directly before any data fusion. We use Ultrasonic and Eddy current sensor data for analysis, both of which are widely used methods of nondestructive inspection for corrosion detection. An efficient method for detecting "region of interest" or ROI for corrosion evaluation and classification is also presented.