Ball screw pair is the core part of NC machine feed system

it is of great significance to study its fault diagnosis. Aiming at the problems of steady-state signal extraction of NC machine tools due to the coupling of operating conditions and processing technology

and the difficulty of data acquisition due to the limited structural space of external sensor installation

an intelligent diagnosis method of ball screw pair based on cross-positional pattern recognition was proposed in this paper. Firstly

a steady-state data extraction method of multi-physical signals based on rotating-speed trend recognition was proposed to quickly obtain steady-state data and eliminate the influence of changes in processing conditions on subsequent analysis. Then

a deep transfer learning method based on improved wavelet convolution is proposed to realize fault diagnosis through cross-location pattern recognition. Finally

the method was verified by using the data of the ball screw pair fault simulation test bench

and the average accuracy of fault identification was more than 97.19%. The research in this paper provides a new method and idea for the research of NC machine tool fault diagnosis and intelligent operation and maintenance under complex working conditions.