Prediction of ultrasonic velocities in ternary oxide glasses using microstructural properties of the constituents as predictor variables; Artificial Neural Network (ANN) approach

Abstract

The longitudinal and shear velocities of ultrasonic waves in glass systems are influenced by the microstructural properties and compositions of the chemical constituents. The relationship between them is highly non-linear and very complex. Artificial Neural Networks (ANN) are adaptive and parallel information processing systems that have the potential to learn by examples and capture the non-linear as well as complex relationships between its inputs and outputs. Neural networks are invaluable where formal analysis would be difficult or impossible. An attempt has been made to predict the ultrasonic velocities in tricomponent oxide glass systems, using the microstructural properties of the constituents as inputs to the ANN.