An Artificial Neural Network Model to Predict Material Characteristics From the Results of Miniature Disk Bending Tests

Abstract

The inverse problem of evaluating mechanical properties of material from the observed values of load and deflection of a miniature disk bending specimen is discussed in this paper. It involves analysis of large amplitude, elastoplastic deformation considering contact and friction. The approach in this work is to first generate—by a finite element (FE) solution—a large database of loaddisplacement (Pw) records for varying material properties. An artificial neural network (ANN) is trained with some of these data. The errors in the various values of the parameters during testing with additional known data were found to be reasonably small.