Modelling of hardfacing layers deposition parameters using robust machine learning algorithms

Abstract The study presents a data-driven framework for modelling parameters of hardfacing deposits by GMAW using neural models to estimate the influence process without need creating experimental samples material and detailed measurements. GAS Metal Arc Welding (GMAW) does sometimes create non-homogenous structures in not only deposited material, but also heat-affected zone (HAZ) base material. Those are fully deterministic, so method should account this unpredictable component learn generic structure hardness resulting Artificial networks (ANN) were used model measured any knowledge equations governing process. Robust learning was decrease outliers noise data on performance. proposed relies modification loss function several them compared evaluated as an attempt construct general analysing electric current arc velocity. can robust layers deposition or other welding processes predict properties materials even unseen based data. This is typically metallurgy, it requires further case studies verify its generalisability.