Prediction of Impact Strength of TIG Welded Cr-Mo Steel Using Artificial Neural Networks

Abstract

Welding is a critical and energy-intensive process with significant importance in the manufacturing industry, enabling the creation of joints capable of withstanding diverse loads without failure. Accurate prediction of welding parameters' effects on the thermal cycle and strength of metals during and after welding is essential to ensure the reliability of welds. This study investigates the influence of welding parameters such as welding current, material thickness, number of weld passes, and electrode diameter on the impact strength of Cr-Mo steel bars. Pure tungsten with 2% thoriated Tungsten Inert Gas (TIG) electrodes was used to join the metal sheets autogenously. Artificial neural network (ANN) was used in creating the model that predicts the impact strength of the steel. Sample with welding parameters of 15 mm thickness, 90 A current, 3 weld passes, and Ø2.4 mm electrode size exhibited the highest impact strength. Furthermore, the analysis of variance (ANOVA) results show that the material thickness and number of weld passes contribute significantly to the impact strength of the steel. The ANN model trained by the Levenberg-Marquardt algorithm had an average training dataset root mean square error (RSME) of 4.12%. This study contributes to the reliability and performance of welded joints in various applications.