Research On Surface Quality In Milling Of SiC-p Reinforced Aluminum Metal Matrix Composite

SiCp/Al composites are one of the most widely used particle reinforced metal matrix composites.As a new material,it has excellent comprehensive performance and has been greatly promoted in the fields of aerospace,military and optical precision instruments.At present,the processing method of this material focuses on the cutting process.However,due to the huge difference between the properties of the reinforced particles and the aluminum matrix,the processing of SiCp / Al composite has its unique difficulties.In this paper,the formation mechanism of SiCp/Al composites and the cause of formation of defects,as well as the factors that affect the quality of the machined surface,are discussed from the theoretical point of view,which provides a way to optimize the milling parameters of SiCp / Al composites.Milling parameters directly affect the milling force and the surface quality of the workpiece.In this paper,16 groups of orthogonal milling experiments are processed to reveal the relationship between the machining parameters and the surface quality of the final machined parts.Firstly,the Taguchi algorithm used in data processing has been introduced.The fluctuation of cutting parameters is regarded as the disturbance of the whole production system.This method can analyze the influence of a specific cutting parameter on milling force or surface roughness in a limited number of experiments.By calculating the signal-to-noise ratio and the fluctuation level of the three cutting parameters,the contribution of each parameter to the result is analyzed.Optimize the milling parameters under the three premise of minimizing milling force,minimizing the surface roughness and ensuring the roughness as much as possible on the basis of comprehensive consideration of milling force and material removal rate.Because surface roughness is an important indicator of surface quality,surface roughness is predicted by using multivariate equations,exponential empirical formula and generalized regression neural network.Multivariate regression equations and exponential empirical formulas are two common mathematical models,however,the error of the models established here are large and can not reach the expected value.The generalized regression neural network has good convergence and approximation performance for small samples.After the data are normalized,the appropriate smoothing factor is selected through cross validation method to train the network,and predicts the validated samples.The predicted results are within the error allowed.In the end,a trained network is used to predict the optimization parameters,and finally selects a group of optimal parameters i.