This paper focuses on the undercut failure between end mill and work-piece in CNC machining of free surface, and presents a new methodology to detect the undercut failure based on the Wavelet analysis. By the typical experiments of the undercut failure, the electric current signal of the spindle motor was monitored and analyzed based on the Daubechies Wavelet transform. The Wavelet smoothing of the monitored signal was used to extract the frequency-band feature. The critical cutting power corresponding to the undercut failure was captured and the undercut feature was extracted with the time-frequency tracking capability of Wavelet transform. Finally, the occurring time of the undercut failure was exactly recognized. The experimental results demonstrated that the undercut failure in CNC free-surface milling could be detected with Wavelet analysis.