The　internal　sensor　signals　of　the　numerical　controlled　(NC)　machine　tools　contain　abundant　information　that　associated　with　the　operating　state　and　the　machining　fault.　However,　the　signal　characteristics　extracted　in　the　time/frecquency　domain　miss　its　physical　significance.　This　paper　presents　a　signal　preprocessing　method　in　the　spatial　domain　to　extract　the　physical　meaning　characteristic　of　the　internal　sensor　signals　with　the　varying　duty　operation.The　proposed　method　uses　the　encoder　signal　to　resample　the　other　condition　signals　in　the　spatial　domain　firstly.　Then,　the　signals　are　analyzed　by　the　Fourier　transform　to　get　the　spectrum.　Compared　with　the　traditional　methods,　the　physical　meaning　of　the　signal　can　be　intuitively　identified.　Moreover,　the　characteristics　can　be　obtained　in　the　varying　duty　operation,　instead　of　the　uniform　motion　in　the　tradition　methods.　It　is　meaningful　for　the　on-line　monitoring,　since　the　working　condition　of　the　machine　tool　is　always　changing　in　the　machining　process.　The　simulations　and　experiments　verify　the　effectiveness　of　the　proposed　algorithm.　©　2013　IEEE.