Modeling of weld height in gas metal arc welding process in the presence of TiO2 Nano-Particles using artificial neural network

One of the quality characteristics of welded joints in gas metal arc welding (GMAW) is weld height (WH). This paper highlights an experimental study carried out to develop a model using artificial neural network (ANN), to predict WH in GMAW in the presence of TiO2 nano-particles. For developing the model, the arc voltage, welding current, welding speed, percentage of Ar in Ar-CO2 mixture and thickness of TiO2 nano-particles were considered as input parameters and WBH as the response. A Doehlert design matrix was employed in the experiments to generate experimental data. The ANN model was developed and validated by conducting five extra runs. The remarkable outcome of this study is the mechanism of arc constriction due to interacting effects between welding input parameters and TiO2 nano-particles. Moreover, the results showed that increasing thickness of TiO2 nano-particles up to almost 0.9 mm increased weld height whereas, its further increase up to 1.0 mm, decreased weld height subsequently. In fact, this variation in weld height could be due to thermal dissociation of TiO2 nano-particles and CO2 releasing oxygen onto weld pool surface, which influenced its surface tension and consequently, changed direction of the Marangoni convection of fluid flow in weld pool and as a result, affected WH. For ANN technique, MSEtrain=0.0066, MSEvalidation=0.0063 and MSEtest=0.0093. Finally, it is to be concluded that ANN is an accurate technique for predicting weld height.