This paper presents an on-line expulsion detection and estimation algorithm for resistance spot welding. Electrode force signal is used to characterize the expulsion signatures. The induced noise is eliminated through an adaptive noise cancellation (ANC) technique. The on-line detection and estimation algorithm is developed based on a generalized likelihood ratio test (GLRT). Experimental results have shown that a 100% detection rate can be achieved for the tests conducted. It is also shown that the energy of the estimated expulsion signature can be used as an index of the expulsion severity.
