China’s manufacturing industry,particularly aerospace manufacturing,relies on aluminum alloy as a primary material.However,the low local stiffness and hardness of aluminum alloy often lead to deformation during cutting,impacting surface quality and reliability.Thus,developing an accurate surface integrity prediction model and optimizing processing parameters pose a challenge in hard aluminum alloy cutting.Currently,empirical methods and swarm intelligence algorithms are used to determine processing parameters.Empirical methods are subjective and non-repeatable,relying on engineer experience and knowledge.Swarm intelligence algorithms,based on optimization theory and computer science,enhance computing efficiency and accuracy but require manual screening and optimization.Therefore,exploring intelligent methods,such as artificial intelligence and machine learning,is crucial to automate processing parameter determination and enhance production efficiency and processing quality.To address these issues,this thesis proposes a multi-algorithm fusion-based method for the intelligent determination of machining process parameters to achieve the minimum surface roughness(Ra),minimum feed direction residual stress(Rsf),and minimum transverse residual stress(Rst)of a workpiece in five-axis milling,enhancing surface integrity and service performance.The main contribution of this thesis includes:(1)Designing four sets of milling machining tests to collect sufficient experimental data within a limited time frame,and using different machining machines for the four sets of tests to improve the robustness of the prediction model.The surface integrity of the workpiece was measured using a measuring instrument after the milling process was completed,and the measured surface integrity was used as the target value and the process parameters selected prior to machining were used as the feature values to form the data set for this thesis.(2)Constructing a surface integrity prediction model using the milling data set and the Stacking integrated learning method to achieve accurate prediction of the surface integrity of the workpiece.The input features of the model are the milling process parameters,including tool attitude(lead angle and tilt angle),tool geometry parameters(tool diameter,number of tool edges,and helix angle),and cutting dosage(depth of cut,spindle speed,and feed rate),and the output prediction values of the model are three indicators of surface integrity,including surface roughness,feed direction residual stress,and transverse residual stress.(3)Conducting multi-objective optimization of the milling process parameters using the non-dominated genetic algorithm NSGA-III to improve the uniformity of the Pareto front and obtain a series of non-dominated process parameters.(4)Employing principal component analysis to determine the optimal process parameters from the Pareto front and verifying the effectiveness of the method through comparative experiments using the entropy weight method.Additionally,two sets of validation tests were conducted on the identified process parameters,and the average error rate for the two sets of tests was 3.9% in surface roughness,and 6.1% and 6.4% in feed direction residual stress and transverse residual stress,respectively.This research provides a new method and idea for predicting surface integrity and optimizing process parameters during milling,contributing to the intelligence of industrial production and playing an important role in industrial applications.Furthermore,the proposed method can effectively improve the cutting surface quality of the workpiece,thus enhancing the service performance of the part. |