Theoretical And Experimental Research On The Surface Microfabrication By Mask Electrolyte Jet Machining Based On Multi-physical Model And Machine Learning Model

Surfaces with microstructures show distinct characteristics in surface energy,sealing properties,wetting properties,mechanical properties,optical properties,thermal properties,fluid dynamics and frictional properties,and have great application prospects in aerospace,aerospace and many other engineering fields.The manufacturing technology of surface microstructure is the core and foundation of surface structure technology.In this paper,mask electrolyte jet machining which combines the electrolyte jet machining and lithography mask is proposed for micro fabrication.Theoretical analysis and experimental results are listed as follows.A multiphysical model of surface microstructure by the mask electrolyte jet machining is established.1.Electrolyte jet morphology based on the Level-Set equation to track the interface and validation with numerical simulations and high-speed cameras.The problem of non-convergence of calculation caused by mesh deformation in the calculation process of large-scale displacement in FEM is solved by using the continuous boundary condition.With the electrode dynamics equation,a multiphysical model based on the mass transfer,electrolyte flow field and electric field is established and verified by experiments.Carried out research on the processing of micro-dimples,micro-protrusions and micro-letters on the surface by MEJM.The consistency of processing size was studied by the nozzle scanning direction at different angles.In the fabrication of complex surface microstructures,the"similarity ratio" index is proposed,and the relative dimensional accuracy is examined.Based on the algorithm of kernel density estimation,the accuracy and repeatability of large-scale processing are studied.Using a different machine learning model for geometry prediction of micro dimples,the signals collected during processing are combined with process parameters to predict processing results.A "pyramid mean" algorithm is proposed to standardize the sampling of signals of different lengths,which solves the problem of feature fusion of signals of different time scales.Different machine learning algorithms are used,and the prediction accuracy between the algorithms is compared.The predictive accuracy of multiphysical model and machine learning model in machining size is compared.