For　the　problem　of　the　workpiece　waviness,　the　relationship　between　the　spindle　unbalance　and　waviness　phenomenon　is　researched.　The　measured　result　of　the　workpiece　is　decomposed　and　restructured　by　the　wavelet　transform,　then,　the　power　spectral　density　is　used　to　analyze　the　spectrum　energy　of　the　signal　in　each　scale　decomposed　by　wavelet　transform,　and　the　corresponding　frequency　is　obtained.　The　modal　information　of　the　spindle　system　is　calculated　by　two　methods.　The　main　errors　features　is　extracted　with　the　frequency　information　of　the　spindle　system,　it　includes　the　natural　frequency　and　rotating　frequency　of　the　spindle,　and　the　spectrum　of　the　alternating　current　interference　of　the　motor,　the　basic　power　frequency　is　50　Hz　and　its　multiple　integers,　sometimes　also　includes　sub-harmonic　signal.　The　extracted　features　from　the　workpiece　is　agree　with　the　unbalance　frequency　of　the　spindle　system　and　the　alternating　current　interference　of　the　motor,　this　explains　the　spindle　unbalance　is　the　main　reason　for　the　waviness,　and　provide　a　identification　basis　for　the　improvement　of　the　machining　accuracy.　©2013　Journal　of　Mechanical　Engineering.