Multi-feature body part programming is difficult for programmers in four-axis and five-axis programming because of their super-many feature relationships and complex vector control.Winding axis programming is a novel solution proposed in multi-axis programming to deal with the problems of super-complex parts with multiple features and difficulty in programming.It can effectively solve the problems of slow efficiency and low fault tolerance encountered in the multi-axis programming of the part.The idea is to process personalized customization as a medium,combining the advantages and disadvantages of three-dimensional programming and planar programming,for complex parts,by reducing the spatial dimension of the programming contour curve,programming is done in low dimensions;the programmed tool path is turned into a curve,then the curve dimension is up,a multi-axis tool path is compiled based on the curve.This method can well solve the problem of multi-axis programming of multi-feature parts,and is an effective programming idea.However,the conversion mechanism is still a bit cumbersome because of repeated dimensionality upgrading and dimensionality reductions.It is also difficult to express with ordinary mathematical formulas or matrices,which is not conducive to in-depth research and rapid application development.This programming method was described.Combined with MATLAB s BP neural network algorithm,the input and output layer and the hidden layer were constructed,and the hidden relationships among them were trained,so as to achieve the purpose of removing the cumbersome transformations in the middle and being able to use them directly.It provided a complete idea and theoretical support for the quick and convenient use of this programming method.Through data analysis,the results of common method programming,dimensionality reduction programming,and MATLAB automatic curve programming were spatially fitted,the errors between the latter two sets of data and ordinary programming results were compared to prove the method feasibility.