NEURAL NETWORK APPROACH FOR FAILURE ANALYSIS OF RECTANGULAR PLATES UNDER WEDGE IMPACT

Abstract

THE PURPOSE OF THIS WORK IS TO ESTABLISH AN EMPIRICAL RELATIONSHIP THAT DESCRIBES THE DEFLECTION CREATED IN A RECTANGULAR PLATE STRUCK BY A RIGID WEDGE AT THE PLATE CENTRE WITH SU±CIENT INITIAL KINETIC ENERGY TO PRODUCE LARGE INELASTIC DEFORMATIONS. A MULTIVARIABLE POWER SERIES WAS SELECTED AS THE FORM OF THE MATHEMATICAL MODEL TO DEVELOP THIS EMPIRICAL RELATIONSHIP. GOOD AGREEMENT BETWEEN THE EXPERIMENTAL RESULTS AND THE PREDICTION OF MAXIMUM DEFLECTIONS FOR VARIOUS IMPACT ENERGIES HAS BEEN OBTAINED. THE DATA USED IN THE DEVELOPMENT OF STATISTICAL MODELS WAS REANALYZED FOR THE PREDICTION OF MAXIMUM DEFLECTION BY EMPLOYING THE TECHNIQUE OF NEURAL NETWORKS WITH A VIEW TOWARDS SEEING IF BETTER PREDICTIONS ARE POSSIBLE. NEURAL NETWORKS HAVE ADVANTAGES OVER STATISTICAL MODELS LIKE THEIR DATA-DRIVEN NATURE, MODEL-FREE FORM OF PREDICTIONS, AND TOLERANCE TO DATA ERRORS. THE NEURAL NETWORK MODELS RESULTED IN VERY LOW ERRORS AND HIGH CORRELATION COE±CIENTS AS COMPARED TO THE REGRESSION BASED MODELS.