Implementation of robotics into welding is an important step toward higher productivity and better quality control of the fabrication process. However most robots currently perform welding in a "blind" fashion. In order to enhance the intelligence of the robots several sensing techniques such as laser stripping, through-the-arc sensing and infrared thermography have been investigated. Most of these sensing techniques are capable of monitoring only a single welding parameter. However infrared thermography has shown promise to detect several types of impending weld defects. 1,2 The results presented in this paper identify approaches to obtain quantitative relationships to monitor the torch position with respect to the seam of the plates being welded. The asymmetry of the thermal profiles caused by arc misalignment has been quantified into a torch seam error relationship. Two principal comparison techniques have been identified. Averaging methods have been implemented to reduce the noise level in the error signals.