Machine learning algorithm comparison for IoT enabled Submerged Arc Welding Process monitoring

Abstract Traditional weld quality monitoring is replaced by IoT enabled remote automation, which reduces rejection and cost due to defective welded products. Multi-sensory based IoT enabled submerged arc wedding process monitoring helps to track live process parameters and scope for data analytics resulting in optimized weld quality. In present research article IoT enabled SAW process monitoring system is used to record process parameter with relevant weld bead characteristic and correlated for desired weld performance and quality. Process parameters of welding region is transferred to IoT enabled user-friendly mobile and computer dashboard using developed with the help of open source MQTT protocol and HIVEMQ platform. Supervised learning algorithm for weld penetration, weld bead width and reinforcement height are analysed with combination of arc current, arc voltage, wire feed rate and weld speed variations. To meet different customer requirement accuracy level of prediction is tested using optimum number of epoch to learns the patterns of input is performed for adaptive weld control strategies for given data set. Prediction of weld bead dimension are carried out using Supervised Machine Learning (SML) algorithms such as Linear regression, Random forest and Decision tree are compared from externally supplied instances to produce general hypotheses, which ca be useful for development of intelligent systems. Random forest algorithm evaluated highest level of accuracy of prediction and lowest mean absolute error for given data set of IoT enabled SAW process data set.