Significant index and adaptive algorithm are necessary for recognizing status of milling cutter. The cutting-force-signal wave was decomposed into valley and amplitude, and formulas were deduced to represent their changes when a flute is broken, in feed direction, the ratio of amplitude value to valley value increases, while that in normal direction decreases. Force power spectrum at tooth rotational frequency and that at spindle rotational frequency represents amplitude value and valley value individually, so their ratios in feed and normal directions were taken as signal features, meanwhile, linear machine algorithm was improved and used, with which the experimental milling-tool data are well clustered, this result shows that the index and algorithm are effective