Prediction of pulsed-laser powder deposits’ shape profiles using a back-propagation artificial neural network

Abstract

The cross-sectional shape profile geometry of blown powder clad deposit is important for overall structural integrity of the clad layer. The penetration zone of the clad deposit is represented within its shape profile geometry. The shape profile geometry of the clad deposit is also important for thermomechanical modelling. Although a rough estimation of blown powder clad deposit shape profile can be made based on the powder feeding parameters using empirical formulae, it may not be sufficiently accurate to be used in a thermomechanical model, as it may lead to inaccuracy in the prediction of temperature distributions, residual stresses, and distortion. The pulsed-laser blown powder deposition process is a highly coupled multivariable problem. Hence the deterministic numerical-methods-based prediction of pulsed-laser powder deposit shape profile geometry is time consuming, costly, and may not be adequate to predict the profile geometry over a wide range of varying process parameters. The present investigation deals with the cross-sectional shape profile geometry modelling of the pulsed-laser assisted superalloy powder deposition (PLPD) process using a soft computing approach. A simple yet effective mapping technique was used in the present work to map the experimentally obtained shape profiles of the powder deposits. The mapped characteristics of the powder deposits’ shape profiles were used in the back-propagation artificial neural network (ANN) modelling of the PLPD process. The present modelling technique can be conveniently used to incorporate the PLPD shape profile geometry parameters in thermomechanical analyses for accurate prediction of temperature distributions and residual stresses. Based on the present soft computing modelling methodology, an estimation of top reinforcement and penetration zone shape boundaries of a blown powder clad deposit can also be made.