<p>Film cooling holes on a thermal barrier coating (TBC) covered blade can be processed by laser-EDM hybrid machining, in which laser pre-drilling is firstly deployed to remove the TBC, and then EDM drilling is used to make a through hole in nickel-based alloys. The depth difference after penetrating the TBC during laser pre-drilling has a significant impact on the subsequent EDM machining. However, inaccuracy in estimating the depth can potentially cause damage to the TBC. Therefore, real-time monitoring of laser drilling and identification of TBC break-out can effectively ensure coating quality. However, due to the uneven thickness of TBC and the complexity of a picosecond laser drilling process, direct identification of TBC break-out is challenging. This paper proposes a qualitative analysis of picosecond laser drilling stages and an on-line identification method for TBC break-out. Based on the evolution of prefabricated holes and detected acoustic signals, a laser drilling process can be divided into three drilling stages and two critical transition instants. The correlation between acoustic signals and drilling stages is established by the time–frequency analysis of acoustic waveforms. Based on the periodicity and frequency domain characteristics of acoustic signals, a classification model for different stages of laser drilling is established using the support vector machine (SVM) method. The generalization performance of the classification model is verified by changing the laser scanning trajectory, processing angle, and laser process parameters. The classification accuracy rate can reach as high as 97%. This paper provides an effective solution to the problem of TBC damages encountered in laser-EDM hybrid machining.</p>