Detecting chatter mark vibration in rolling mills operating under normal working conditions is difficult. A novel characteristic recognition method of chatter mark vibration, where the non-dimensional parameters are calculated with time varying signals and kurtosis under normal rolling mill operating conditions, is presented in this paper. The character of the chatter mark vibration signal is obtained by calculating the kurtosis value of each vibration signal segment obtained by subdividing the raw time varying vibration signal. The probability density function is utilized to reveal obvious differences between signals with respect to the normal and chatter conditions. The method overcomes the limitation of traditional spectrum analysis, which is sensitive to working conditions. Numerical simulation and experimental results show that the proposed method has better recognition capability than traditional spectrum analysis.