NDT of Piled Foundations: Data Processing with Artificial Neural Networks

Abstract

The low strain test method is the prevalent method for integrity testing of cast in situ foundation piles. The automated interpretation of sonic echo traces from this test may prove beneficial to industry through standardisation of the test method and a reduction in the time spent analysing data. In this research the generalisation and feature extraction strengths of artificial neural networks have been exploited for test data interpretation. This study involved the use of a multilayer perceptron network considered most suitable for this heteroassociative function approximation task. The network was trained using numerically generated data. Field data from two sites confirmed that the network identified, located and quantified changes in pile diameter as well as detecting pile lengths to within 5% of their design values.

Wavelet transform and least square support vector machine for mechanical fault detection of an alternator using vibration signal

“…Wavelet coefficients cannot be directly used as inputs to a classifier, so a pre-processing stage is essential to prepare data for the classifier. 15 Feature extraction 16 is used in this research for determination of useful data from vibration signals of the alternator. The extracted features are used as inputs to the classifier for signal identification and fault classification.…”