Abstract: The importance of welding quality in metal production cannot be overstated because it improves the durability, toughness, and strength of engineering structures. The assessment of weld quality involves various parameters. Traditional methods such as welder expertise, charts, and handbooks have been used to determine desired welding parameters, offering simplicity and cost-effectiveness. However, relying solely on these methods doesn't guarantee satisfactory welding outcomes, especially in new welding processes. To address this challenge, the study aims to utilize artificial intelligence models for parameter optimization. The mild steel plate was chosen as the research material due to its availability. An optimal experimental design was carried out using design software. Gas tungsten arc welding was employed to create weld samples, with input factors; gas flow rate, voltage, and current. The desired outputs were the weld strength factor, weld factor of safety, and weld quality index. Both the response surface methodology (RSM) and artificial neural network (ANN) models were utilized to generate optimal solutions for controlling and predicting experimental responses. The RSM model was developed, tested, and validated, demonstrating high strength and accuracy in maximizing weld strength, quality index, and weld factor of safety. Similarly, the ANN model provided close correlations with experimental results, enhancing prediction capabilities.
Keywords: Design of experiment, Response Surface Methodology (RSM), Artificial Neural Network (ANN), weld strength factor.