Multi-objective optimization based IBCS for surface roughness and textural feature in MCVE piston machining

In this paper, the effects and multi-objective optimization of process parameters on surface roughness (R a ) and textural feature in middle-convex and varying ellipse (MCVE) Al-Si alloy piston machining are analyzed. The indicator of textural feature is the variance of textural width (V t w ), which could estimate the piston skirt lubrication. The cutting force model is proposed to investigate the machining motion and turning force in single cross-section of piston showing that the turning of MCVE piston is more complex than ordinary turning. In this turning process, the key process parameters such as feed speed, spindle speed, tool elongation, and tool tip radius are optimized. Experimental data were initially collected based on the Taguchi method of experimental design, which is L25 (53) Taguchi standard orthogonal array on cutting experiments. From all of these parameters, the main effect of tool tip radius process parameter on R a or V t w is more obvious. However, the performance of R a and that of V t w is contradictory in single-objective optimization for the process parameter. By analysis of contour plot, the regions of optimal levels are small and dispersed for V t w . Hence, it is difficult to obtain optimal levels of process parameters. Under simulation, the alternation effect of cutting forces is more pronounced. Additionally, the suitable performance of V t w will be obtained while the tool tip radius (α) is small. Finally, an integrated system is proposed based on multi-optimization techniques, such as regression models, improved beta-distribution cuckoo search (IBCS) algorithm and desirability function. It shows that the average number of iterations of IBCS to achieve the optimal solution is approximately 10, which is better than that of beta-distribution cuckoo search (BCS) or cuckoo search (CS) algorithm. This result could be attributable to the fact that the parameter α0 in Levy Flights is updated by the beta-distribution of exponential decay in the proposed IBCS algorithm. Confirming experiment shows that the mean prediction error of all is no more than 15%.