Research On Chatter Diagnosis And Multi-objective Parameter Optimization In Milling Process

Aluminum alloy materials are widely used in the processing and manufacturing of aeroengine blades,blower impellers,automobile cylinders,etc.,and the flutter problem in the milling process seriously restricts the improvement of production efficiency and quality.Research on the diagnostic methods of the milling process is significant.In addition,how to achieve multi-objective parameter optimization under multiple constraints is also a hot topic in the field of manufacturing.Aiming at the problems and research hotspots in the above aluminum alloy milling process,a milling force model for ball-end milling cutters is established in this paper.A two-degree-offreedom milling dynamics model based on the completely discrete method and an SCA multiobjective parameter optimization model are explored to explore the milling force signals.Feature extraction of flutter in the time-frequency domain,and the application of BP neural network for flutter diagnosis research.The milling force model can be used to predict the milling force and calculate the maximum instantaneous force.The dynamic model of the toolworkpiece system is convenient for exploring the effects of different milling parameters under stable conditions and designing milling experiments from stable to chatter.The main research contents of this topic are as follows:(1)The milling force model of the ball-end milling cutter is established by discreteintegral-sum method of the tool geometry,and the milling force coefficient is determined by the variable feed slot milling experiment method.The modal parameters of the tool,workpiece and tool-workpiece system were determined by a combination of finite element simulation and hammering experiments,and a stability prediction model based on the fully discrete method was established.The influence of different milling parameters on the milling force was explored under the guidance of the stability model.(2)The main signal processing method used for flutter feature extraction is variational modal decomposition.The theoretical knowledge related to variational modal decomposition is introduced,including the model construction and solution process.This paper explores the important parameters of variational modal decomposition—the number of modal K and the influence of penalty factors on it.The introduction of GWO solves the problem of automatic determination of two parameters of variational modal decomposition.Finally,a parameter optimization variational modal decomposition theory based on Lempel-Ziv complexity is proposed.(3)Based on the theoretical knowledge of flutter feature extraction,the time-domain variance feature and frequency-domain energy feature are added as the feature input of flutter diagnosis.A flutter diagnosis theory based on BP neural network is used to build a flutter diagnosis system The experimental platform also designed the software of flutter diagnosis system.(4)A three-dimensional micro-surface topography model was established,and the effects of milling parameters and tool parameters on them were explored.A surface roughness prediction model based on the traditional linear regression method and GWO-LSSVM method was established,and the prediction accuracy' and efficiency of the two were compared.Surface morphology observation experiments from stabilization to flutter were performed.A multiobjective optimization model based on SCA is established and a case analysis is performed.The optimization performance of multiple optimization algorithms for the same model is also compared.