Correlation of welding current waveform with welding conditionbased on exploratory data analysis (EDA)

• Main Author: Afidatusshimah, Mazlan
• Format: Thesis
• Language: English
• Published: 2019
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://umpir.ump.edu.my/id/eprint/27968/1/Correlation%20of%20welding%20current%20waveform%20with%20welding%20conditionbased%20on%20exploratory%20data%20analysis%20%28EDA%29.wm.pdf; http://umpir.ump.edu.my/id/eprint/27968/

Welding is a fabrication method that is used worldwide, especially in the manufacturing and automotive industries. The conditions of the welding weldment are important to ensure the quality of the product. Hence, the quality of the product is a big concern and strict requirement to deliver a good production. Therefore, the nondestructive test like dye penetrant test, ultrasound test, and radiographic test plays the role in certifying the quality of weldment based on ASME standard. Unfortunately, the nondestructive tests are costly and time consuming. Thus, a real-time monitoring method was applied in this study. The real-time monitoring allows the test to be carried out during welding activities where the results can be obtained immediately. In previous researches, studies on realtime monitoring used the welding signal such as arc light signal, sound welding, and welding current. Among the welding signal, welding current was chosen because of its simple signals, easy to collect, and rich with information in welding process. This study was carried out by welding current signal and welding condition monitoring and produced two results. Then, both results were proceeded to manual syncing and sliced into 1 mm pieces data for high sampling data. When more data are collected, they are more precise and have more resolution. Among the current characteristics, the current’s peak count is the most influential variable to correlate with the welding condition. Based on the current’s peak count, the good and defect conditions can be distinguished. Next, the welding condition data and the welding current pattern were analysed using exploratory data analysis (EDA) and the findings were concluded in this study. The analysis shows that the results support the earlier findings. In this experiment, metal inert gas (MIG) welding was used and set up at the Faculty of Mechanical Engineering, Universiti Malaysia Pahang in room temperature. The outputs of the experiment were the welding sample and welding current. The conditions of welding samples were identified by a qualified person in welding. Based on the result, the welding condition and welding current were compared using manual syncing of the length waveform. Then, the data were sliced into 1 mm data and analysed on EDA. In the end, the analysis shows that there was a significant difference between the welding samples in good condition and welding sample in defect condition using current’s peak count variable. This variable indicated similarities and differences between welding sample in good condition and defect condition. From the experiment, 8 out of 10 defect conditions were likely to be detected by examining the current’s peak count compared to the good welding condition. Among the defects, incomplete weld and lack of penetration (LOP) defects show differences in the current’s peak count whereas similar current peak count was found among other defects. As a conclusion, the welding current’s peak count can identify the conditions of welding sample whether it is in good or defect condition. In the future studies, the research can be improved by exploring each of the defect types based on the current pattern with different equipment and types of metal.