Abstract: The detection of effective ultrasonic echo about multi-component is the key point to realizing coal-rock inter-ference recognition. In actual ultrasonic detection,as the echo received is the signal composed of multi-component which is low SNR,non-stationary and non-linear,it’ s hard to accurately and efficiently get the waveform about the coal-rock boundary. Traditional wavelet transform method requires that the specific wavelet base is selected according to the characteristics of signals. Frequency aliasing and illusive components exist in adaptive decomposition methods such as EMD,EEMD etc. VMD (Variational Mode Decomposition) is proposed to analyze the components of the ultra-sonic echo signal and precisely obtain the waveform and time of the first echo signal about coal-rock boundary so as to realize the interface recognition among different components as well as make the judgement of related parameters. Sim-ulation and experiments demonstrate that VMD,able to achieve a more optimized signal decomposition effect,makes a difference in accurately identifying the interface of multi-components as represented by coal-rock.