Based on the study and analysis that Lyapunov exponent can be used to characterize the motion state of system, in connection with the nonlinear dynamic characteristics shown from the performance of fault rotating mechanical system, the authors proposed the analysis method of Lyapunov dimension on the signal feature of mechanical fault. Using theory of phase space reconstruction, simulating fault signal of rotating machine was reconstructed. In order to let the phase space reconstruction adequately reflect the movement characteristics of the system, the time delay and embedding dimension are discussed emphatically. Based on what mentioned above, the authors calculated the Lyapunov exponent and the Lyapunov dimension. From the analysis and calculation on simulation of different fault signals, it showed that under different rotating machinery fault conditions, its Lyapunov dimension were significantly different, which verified that this nonlinear feature quantities were effective parameters for fault information and they were excellent parameters in terms of extraction and recognition of fault features.