<p>A novel method based on multifractal theory was proposed to identify grinding chatter, based on the discovery that the multifractal spectrum width of the vibration signal under stable grinding is obviously larger than that under chatter grinding. The original vibration signal was decomposed into intrinsic mode functions (IMFs) covering different frequency bands with the variational mode decomposition (VMD) method. The multifractal spectrum width of each intrinsic mode function was calculated, revealing that the intrinsic mode function IMF-4 in rail grinding exhibits the most distinguished multifractal spectrum width during chatter from stable grinding, in accordance with the fact that IMF-4 covers the chatter frequency. The multifractal spectrum width of IMF-4 was selected as the indicator of chatter in rail grinding and validated with 40 tests under different grinding conditions, with apparently different thresholds in stable grinding from those in chatter grinding.</p>