Classification of Micro-Damage in Piezoelectric Ceramics Using Machine Learning of Ultrasound Signals

Abstract

Ultrasound based structural health monitoring of piezoelectric material is challenging if a damage changes at a microscale over time. Classifying geometrically similar damages with a difference in diameter as small as 100 μm is difficult using conventional sensing and signal analysis approaches. Here, we use an unconventional ultrasound sensing approach that collects information of the entire bulk of the material and investigate the applicability of machine learning approaches for classifying such similar defects. Our results show that appropriate feature design combined with simple k-nearest neighbor classifier can provide up to 98% classification accuracy even though conventional features for time-series data and a variety of classifiers cannot achieve close to 70% accuracy. The newly proposed hybrid feature, which combines frequency domain information in the form of power spectral density and time domain information in the form of sign of slope change, is a suitable feature for achieving the best classification accuracy on this challenging problem.

AI and BD in Process Industry: A Literature Review with an Operational Perspective

“…Ceramics: ML is applied for quality control in the ceramics industry for the detection and classification of defects in the final product in combination with other technologies, like ultrasound sensing [20]. IS using ML (ANN) for the analysis of acoustic emissions and cutting power signals [21] have been applied to predictive maintenance.…”

Characterization of ceramics based on laser speckle photometry

“…When thermal waves propagate on a normal surface without defects, a homogeneous spatial gradient of temperature is formed. If surface defects exist, the temperature distribution becomes nonhomogeneous; i.e., the presence of defects changes the distribution of thermal energy around the defect area (Tzou, 2014). Heat energy then accumulates around the defective area, as presented in Fig.…”

Learning velocity model for complex media with deep convolutional neural networks

“…The works [1,2] successfully applied deep learning for non-destructive testing data interpretation and defect classification. The approaches presented can work in realtime and provide new options compared with traditional imaging and interpretation methods.…”