Automated Monitoring of Bonding Materials' Properties in Complex Structures using Machine Learning

A. Chlingaryan (Armenia) and N.S. Melkoumian (Australia)


Machine Learning, Acoustic Emission, Simulation, Complex Structure


A method is proposed for automated monitoring the properties of bonding materials in complex structured. An acoustic source is used to generate waves in the structure which are then registered by a sensor located on the other side of the bonding interface. The influence of the bonding material on the onset time is used to predict the elastic modulus of the bonding material. The proposed method is based on statistical supervised machine learning. The correlation between the onset time and the elastic properties of the bonding material are modelled employing Gaussian processes. The proposed approach is cheap to implement, can automatically clean the noise in the dataset and with small training dataset can produce reliable non-linear probabilistic model for predicting the properties of the bonding material using the onset time of the acoustic waves.

Important Links:

Go Back