Time of Flight Measurements in Real-Time Ultrasound Signatures of Aluminum Spot Welds: An Image Processing Approach

Ultrasonic testing is one of the most popular non-destructive methods to determine the quality in spot welds. The next step of this technology is to be able to determine the quality while the weld is being made, this advance can have a great impact on the quality assurance of joints in materials like aluminum, where the spot welding process is a difficult task. In our approach, a state-of-the-art ultrasound transducer assembly is installed in one of the welding electrodes. Working in pulse-echo mode, the system collects A-scans of the waves passing through the welded plates in the direction perpendicular to the plate’s surface. A-scans are gated to allow stack-up front and back wall reflections to be recorded. Such setup has allowed the acquisition of real-time ultrasonic signatures composed of multiple A-scans for each spot weld. Eventually the signature should be processed to extract information about the weld quality. During the welding process temperature gradually increases within the aluminum plates; as sound speed in metals is inversely proportional to the temperature, the time of flight will increase during welding. The change in time of flight of the ultrasonic wave during the whole welding process is a good indicator of the heating rate in the weld and can be a good judge of weld quality. Unfortunately, electrode deterioration, industrial noises and pulse changes due to frequency dependent attenuation make signal detection a challenging task. To overcome this difficulty the system treats B-scan as an image instead of a set of separate A-scans. With such approach, the weak signal detection (in most of the cases, weak back wall reflection) is performed by applying different image processing techniques which take advantage of statistical properties of neighbouring A-scans (signal similarity). This paper presents an effective and robust means of monitoring the change in time of flight on several signals acquired during the welding process. This is achieved by an image processing technique which exploits the B-Scan representation of the ultrasonic signature of the weld. The proposed method is able to detect the back wall reflection during the complete welding process with sub-pixel resolution and is able to interpolate areas where the signal is not present.