Feasibility study of in-process weld quality control by means of scanning laser profilometry

Abstract

We have developed a scanning laser system based on the optical triangulation principle for the shape measurement of fusion weld surfaces. The system integrates a triangulation module consisting of a laser line projector and a digital video camera with a mechanical scanning stage and an industrial computer. The system is small and rugged, suitable for application in industrial environment. The system can sample a weld surface at a rate of up to 30 profiles per second achieving 0.1mm accuracy. Software was developed which analyses the captured weld surface shape in real time determining the characteristic shape parameters (length, width, height, cross section, volume, starting position,...) which are then used for automated classification of the welds into acceptable and unacceptable. The software also detects surface defects such as undercutting, holes or melt splash, etc. The system has been tested in a robotized welding cell for automotive parts in an industrial production facility. Weld classification obtained by the system was compared to an independent classification determined by a trained weld inspector on the basis of visual inspection and to another one determined on the basis of metallographic analysis of the weld. Using the metallographic based classification as the reference we find that the developed weld inspection system can achieve better classification reliability than a trained visual weld inspector.