Machining-induced residual stress analysis and multi-objective optimization for milling process of Mg–Li alloy

Abstract

Machining-induced residual stress in an Mg–Li alloy can lead to a significant deformation of its machined thin-walled parts. To solve this issue, an integrated method that combined the grey relational analysis (GRA), back propagation (BP) neural network, and non-dominated sorting genetic algorithm-III (NSGA-III) was proposed in this study aimed to analyze the dominant mechanism of residual stress and achieve a super-objective cooperative optimization of residual stress in the LA103Z Mg–Li alloy milling process. A typical spoon-shaped distribution of surface tensile stress and subsurface compressive stress was presented in the Mg–Li alloy owing to the coupling interaction between the cutting force and temperature. The increase in the cutting temperature owing to an increase in the cutting speed was proven to be the overarching factor that induced residual stress by the correlation analysis of cutting parameters, force–heat of cut and residual stress. The BP prediction models of the residual stress were established, and the predicted values were in good agreement with the test values, with a maximum error of 5.07 %. The NSGA-III optimization algorithm was used to search for the lowest values of in-plane tensile and compressive stresses (σtx, σty, σcx and σcy). The optimized cutting speed, feed, and cutting depth were determined to be 61 m/min, 0.068 mm/z, and 0.58 mm, respectively. The maximum predictive error of four residual stresses was experimentally verified to be 15.94 %, which proved the feasibility of the multi-objective optimization approach. The residual stress level of the optimized scheme evidently decreased compared with that of the current milling scheme, with the maximum reduction of 28.37 %. Therefore, the propsoed approach can effectively reduce the risk of warping and distortion of machined Mg–Li alloy parts.