It is very important to track a complicate weld seam for the welding automation. Very recently, laser vision sensor becomes a useful sensing tool to find the seams. Until now, however studies of welding automation using a laser vision sensor, focused on either image processing or feature recognition from CCD camera. Even though it is possible to use a simple algorithm for tracking a simple seam, it is extremely difficult to develop a seam-tracking algorithm when the seam is more complex. To overcome these difficulties, this study introduces a simulation system to develop the seam tracking algorithm. This method was verified experimentally to reduce the time and effort to develop the seam tracking algorithm, and to implement the sensing device.