Date of Award


Publication Type

Doctoral Thesis

Degree Name



Electrical and Computer Engineering

First Advisor

Roman Gr. Maev


Electrical engineering




In this work, ultrasound is used as a non-destructive method of monitoring the welding process in real-time to detect expulsion events. During spot welding, a single element ultrasound transducer placed behind one of the welding electrodes operates in pulse-echo mode and probes the axial center of the welded zone. Acoustic reflections from the electrodes, plate interfaces and liquid metal weld nugget are recorded as A-scans. During welding, the A-Scan reflections change with time, since the material properties of steel (e.g. density and elasticity) change with temperature. Imaging successive A-scans in time forms an M-Scan image of the welding process from which the dynamic formation of the spot weld can be depicted and analyzed. This thesis focuses on taking a brand new approach to the problem of expulsion detection by identifying and characterizing expulsion events in M-scan data. Expulsion occurs when molten material is ejected from the welded zone as a result of overheating due to: poor electrical/thermal contact, coating thickness and/or excessive weld current. An expulsion can have a significant impact on the final yield strength of the weld, and thus the detection and characterization of expulsion events is significant to the quality assurance of resulting spot welds. The main contribution of this work was the discovery of M-scan features that provide a means of detecting, predicting and classifying the event. These include: 1) Detection by sudden phase delay change of the workpiece surface reflection. 2) Prediction by ultrasonically measuring the heating rate prior to expulsion. 3) Classification of the weld quality by ultrasonically measuring indentation in the heated workpiece. In addition, new methods for automatically detecting and measuring these features were developed that utilize a new efficient Hough transform variant proposed in this work. It was shown using both lab experiments and industrial data that not only does the automatic detection of these features provide a new and robust means of identifying expulsions in a wide range of welding setups, but this research can also be used in the future to provide real-time feedback to dynamic weld controllers and eliminate expulsions from occurring altogether.