Online modeling and prediction of weld bead geometry in robotic gas metal arc based additive manufacturing using grey prediction model

Abstract

In the directed energy deposition process, online monitoring of weld bead geometry can give real-time information to researchers and manufacturers to understand and control the metal deposition layer by layer. In the current work, an optimization Grey model OGM(1,N)-based online monitoring system was built to monitor the height and depth of metal deposition in each layer. The OGM(1,N) needs fewer training samples with poor information. The training data was updated to improve accuracy in prediction by removing old data and introducing more recent data. The inputs to the OGM(1,N) include the welding time, current, absolute difference in current at regular time intervals (5 s) and arc force. An optimum set of input parameters was identified by calculating the root mean square error (RMSE) for various parameter combinations for the most accurate weld bead height and depth prediction. The interaction of time, current, and arc force significantly affected the height and depth of weld beads. The predicted height and depth of weld beads using the time, current, and arc force showed RMSEs of 2.52 and 0.23, respectively, compared to experimental results. This methodology can assist manufacturers and researchers in understanding and achieving an appropriate balance between process parameters and weld bead height with no training time.