The objective of this research is to apply new sensing techniques, artificial intelligence, and robotics to improve the welding process through control of penetration depth and width parameters. By producing a constant depth and width of penetration through on line control, the quality and strength of welds may be improved to a new level of reliability previously unobtainable. Experimental data presented in this paper indicate that depth and width of penetration can be monitored by infrared thermography and hence controlled dynamically during the weld process. Infrared thermography is used to monitor surface temperature distributions in the vicinity of the molten metal pool. Characteristics of temperature distributions perpendicular to the direction of arc travel are analyzed and correlated with weld bead width and depth as measured by destructive examination after the weld production. The results show that there is a linear relationship between peak weld plate surface temperature and depth of penetration. Additionally, surface puddle width is found to be directly correlated to measured distance separating solidus temperatures of the molten metal as measured by infrared thermography.