Classification of Corrosion Detected by Acoustic Emission

Abstract

This paper presents an Acoustic Emission (AE) to detect pitting corrosion in stainless steel. The AE signals were analyzed to reveal the correlation between AE parameters and severity levels of pitting corrosion in austenitic stainless steel 304 (SS304). In this work, the corrosion severity is graded roughly into five levels based on the depth of corrosion. Relationships between a number of time-domain AE parameters and the corrosion severity were first studied and key parameters identified. The corrosion severity was also categorized into three stages: initial, propagation and final stages based on the source mechanisms of the AE signals. We identified these stages from the frequency-domain characteristic of the AE signal and the visual characteristic of the corroded pits in each level of corrosion severity. A number of measures were employed to quantify such characteristics and the source mechanisms hypothesized. To demonstrate the usefulness of such parameters, a feed-forward neural network was used to classify the corrosion severity. Preprocessing and verification techniques were provided to facilitate and to maintain the generalization capability of the network. The classification performance is excellent and demonstrates that the AE technique and a neural network can be efficiently used to detect and monitor the occurrence of corrosion as well as to classify the corrosion severity.

Semi-Parametric Learning for Classification of Pitting Corrosion Detected by Acoustic Emission

“…The period of experiment was three hours for each specimen. The data was then undergone a preprocessing step before being fed into classifiers by a linear scaling so that it stayed within the range of [-1,1] and filtered by an Adaptive Moving Average (AMA) [6,7].…”

“…An additional set of prior probabilities was also calculated from the training set for MAP classifier using the amounts of data in classes. For FFNN, a single hiddenlayer of eight nodes as suggested by our previous work [6] was used.…”

“…Corrosion monitoring by Acoustic Emission (AE) has become more active recently [1][2][3][4][5] because it does not require any process interruption. Although the implementation of AE for corrosion monitoring was frequently studied, only few of them [6,7] focused on classification of corrosion severity.…”

“…The purposes of this paper are to confirm the ability of pitting corrosion detection in stainless steel, to analyze the relationship between the obtained AE signal and the corrosion, and to apply practical classification techniques based on Bayesian Statistical Decision Theory [8], namely Maximum A Posteriori (MAP) and Maximum Likelihood (MLE) classifiers for grading the severity level of the corrosion. The result was compared with other classification methods [6,7] to compare its performance. The ultimate goal of this work is to develop a novel corrosion monitoring technique that can be used in maintenance planning works.…”

Acoustic emission corrosion feature extraction and severity prediction using hybrid wavelet packet transform and linear support vector classifier

“…Time domain analysis were performed on AE signals and discovered a link between AE Energy and corrosion. Saenkhum et al [ 6 ] classified corrosion using acoustic emission and an Artificial Neural Network (ANN). Four characteristics experiment-derived AE energy, amplitude, rising time, and count were employed as inputs to a neural network.…”