Optimum Welding Parameters for Pipeline Welding Using Artificial Neural Network

• Main Author: Krishnamoorthy, Shasidaran
• Format: Final Year Project
• Language: English
• Published: IRC 2016
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://utpedia.utp.edu.my/17273/1/FYP%20Shasidaran%20%2818625%29dissertation%20reportdocx.pdf; http://utpedia.utp.edu.my/17273/
• Institution: Universiti Teknologi Petronas

In the process of MIG welding, the welding parameters such as welding current, arc voltage and welding speed has a significant effect to the weld joints mechanical properties and thus affect the quality of the weld from the aspect of mechanical properties. Even small variation in any of the cited parameters may have an important effect on depth of penetration. In this study, stainless steel 316L (316L) were chosen as the base metal to be tested using the main parameters of MIG welding. All the welding procedures were done according to the standards provided by American Welding Society (AWS). Physical properties preferred in any welded components are like tensile strength, yield strength and elongation. To achieve these physical properties, penetration is the key parameter to be verified. The process of mechanical properties testing was done accordance to ASTM E8/E8M standard, to make sure all the methods carried out are valid. Moreover, the welding process was performed using sets of input parameters to obtain specific results which was used to tabulate through mathematical modelling, as a procedure in optimizing the weld parameters, which is the regression model and the data sets were used to train and develop artificial neural network (ANN). In this project, a study on the welding parameters for pipeline was done by application of MIG welding by taking into account welding speed and wire feed rate as the parameters. The parameters are important to determine the tensile strength and weld bead penetration of the welded specimen. The data sets are necessary to train the Neural Network using Matlab ANN tool and hence enable to predict the desired output which was compared to the experimental data to check for validation. Therefore, the ANN predicted results shows a regression value of 0.95664 and 0.90948 for tensile strength and weld bead penetration respectively which means that the predicted value is near to the experimental value for the desired inputs which satisfy the objective of the study.