基于MOWOA算法的42CrMo钢超声滚挤压加工参数优化  

Optimization of ultrasonic rolling extrusion processing parameters of 42CrMo steel based on MOWOA algorithm

作　　者：石青松 徐红玉[1] 王晓强[1] 张旭 SHI Qing-song;XU Hong-yu;WANG Xiao-qiang;ZAHNG Xu(College of Mechanical and Electrical Engineering,Henan University of Science and Technology,Luoyang 471003,China)

机构地区：[1]河南科技大学机电工程学院,河南洛阳471003

出　　处：《塑性工程学报》2024年第8期48-57,共10页Journal of Plasticity Engineering

基　　金：国家自然科学基金资助项目(U1804145);国家重点研发计划(2018YFB2000405,2022YFC2805702)。

摘　　要：为提高42CrMo钢零件的表面质量与抗疲劳性能,设计正交实验分析了超声滚挤压加工参数的显著性及其对表层性能指标的影响规律。基于实验数据建立了BP神经网络与指数回归预测模型,对比验证模型的精确性。对预测模型采用多目标鲸鱼算法(MOWOA)进行了三目标和双目标优化,得到了加工参数和表层性能最优参数集合并分析了表层性能指标之间的权衡关系。结果表明,指数预测模型预测精度更高,最优加工参数集合为:转速210~250 r·min^(-1)、进给速度12~16 mm·min^(-1)、振幅25~28μm、静压力517~630 N;最优表层性能参数集合为:表面粗糙度0.466~0.507μm、残余压应力1002~1110 MPa、显微硬度709~720 HV。实验验证了算法的准确性。To improve the surface quality and fatigue resistance performance of 42CrMo steel parts,the orthogonal experiments were designed to analyze the significance of ultrasonic rolling extrusion processing parameters and their influence laws on surface performance indexes.Based on experimental data,the BP neural network and exponential regression prediction models were established to verify the accuracy of the model.The prediction model was optimized by using multi-objective whale algorithm(MOWOA)to perform three-objective and two-objective optimization,and the sets of processing parameters and surface performance optimal parameters were obtained,and the trade-off relationship between surface performance indicators was analyzed.The results show that the exponential prediction model has higher accuracy and the optimal set of processing parameters is:rotation speed of 210-250 r·min^(-1),feeding speed of 12-16 mm·min^(-1),amplitude of 25-28μm,static pressure of 517-630 N.The optimal set of surface performance parameters is:surface roughness of 0.466-0.507μm,residual compressive stress of 1002-1110 MPa,microhardness of 709-720 HV.The accuracy of the algorithm was verified by experiments.

关 键 词：超声滚挤压 BP神经网络 指数模型 多目标鲸鱼算法 表层性能 

分 类 号：TG376.1[金属学及工艺—金属压力加工]