With the development of large precision machinery including aircraft engines, compressors towards high load, high efficiency and high reliability, the quality requirements of the impeller are getting higher and higher. However, as the processing technology for impeller is complex and the cutting condition is unstable, milling tool edge is easily broken in machining process, seriously affecting the quality of impeller. The relationship between the milling force and eddy-current displacement signal of cutter radial swing is verified as a linear tend. The vibration signals of acceleration sampled from x and y directions of machine tool spindle end are transferred into vibration signal of displacement with frequency-domain integration algorithm. Then, the fundamental and harmonic components of integration signal are extracted for further analysis, which solves the difficulty of eddy current sensor installation and effectively removes influence of the jamming signal. The first-order cumulant and second-order cumulant with time synchronous averaging algorithm and characterize waveform of the cumulants quantitatively with characteristics are calculated in time domain, which include variance, skew, kurtosis, absolute mean value and RMS. At last, setting threshold can realize the state warning. It can be concluded that the proposed method can solve the problem of online monitoring and early warning of milling cutter condition in complex surface machining process.