摘要
目的为提高激光作用于H13钢表面的硬化效果,提出一种基于鱼鹰优化算法(OOA)、优化随机森林算法(RF)的H13钢激光表面硬化工艺参数预测方法。方法首先采用有限元模型粗选激光功率、扫描速度、搭接率等工艺参数的范围,然后在该工艺参数范围下设计三因素五水平的中心复合试验(CCD),通过有限元模型得到硬化层参数。然后,分别构建响应面(RSM)、RF和基于OOA-RF的平均淬透深度、峰谷差值、峰值温度的预测模型,并对3种模型的预测精度进行分析对比。通过多目标遗传算法(NSGA-Ⅱ)进行工艺参数寻优,并结合优劣解距离法(TOPSIS)和熵权法(EWM)对寻优解集重新进行排序,从而得到最佳工艺参数组合。结果在最优工艺参数功率为517W、扫描速度为5mm/s、搭接率为48%下进行试验,得到平均淬透深度为723.3μm,与预测值的相对误差为11.15%,峰谷差值为58.75μm,与预测值的相对误差为3.77%,硬化表面平整,无明显凹陷现象,激光表面硬化前的硬度为(165.2±9.2)HV0.5,经表面硬化后提升至(381.4±86.2)HV0.5,平均硬度提高了约1.3倍。结论可为合金激光表面硬化工艺参数的寻优提供参考。
To enhance the hardening effect of laser action on the surface of H13 steel,this study proposed a method for predicting the process parameters of laser surface hardening based on the Osprey Optimization Algorithm(OOA)optimized Random Forest(RF)algorithm.Firstly,a finite element model was established to simulate the temperature field changes on the surface of the workpiece during laser scanning.Experiments were then conducted under the same process parameters,and the maximum quench depth was measured to validate the effectiveness of the model.The results showed a relative error of 11.0% for the maximum quench depth,indicating that the established finite element model accurately reflected the laser surface hardening process and provided support for selection of process parameter ranges in subsequent studies.The finite element model was utilized to determine the process parameter ranges for laser power,scanning speed,and overlap rate.Subsequently,a three-factor,five-level central composite test(CCD)was conducted within these ranges to derive hardened layer parameters through the finite element model.Then,a response surface methodology(RSM)surface hardening layer prediction model,an RF surface hardening layer prediction model,and a surface hardening layer prediction model based on OOA-RF were constructed separately,and the prediction accuracy of the three models was analyzed and compared.The OOA-RF model was found to have a higher goodness of fit(R^(2))for the response target compared with the RSM and RF models,indicating its better applicability.Additionally,the mean absolute percentage error values of the OOA-RF model were consistently lower than those of the RSM and RF models,further highlighting its higher fitting accuracy for the response target.By using the multi-objective genetic algorithm(NSGA-Ⅱ)to optimize the established OOA-RF model for process parameters,constraints needed to be applied to both the range of process parameters and the response objectives to achieve a good quenching effect.After optimization,the Pareto front solution set was obtained.To select the best solution from the solution set,the Technique for Order of Preference by Similarity to Ideal Solution(TOPSIS)and the Entropy Weight Method(EWM)were combined to re-rank the optimization solution set and obtain the best combination of process parameters.In the verification test conducted at the optimal process parameters with a power of 517 W,a scanning speed of 5 mm/s,and an overlap rate of 48%,a cross section perpendicular to the scanning path was cut.After polishing and etching the cross section,the average hardened depth was observed to be 723.3μm,with a relative error of 11.15% compared with the predicted value.Additionally,a peak-to-valley difference of 58.75μm was achieved with a relative error of 3.77% from the predicted value,and hardened surfaces were relatively flat,without visible depressions.The hardness before laser surface hardening was(165.2±9.2)HV0.5.After surface hardening,the hardness increased to(381.4±86.2)HV0.5,indicating an average hardness enhancement of 1.3 times.The elemental content of hardened and non-hardened areas was similar,and the main reason for hardening was the martensitic transformation of the hardened layer.This method demonstrates potential for optimizing laser surface hardening process parameters for alloys.
作者
梁强
徐彬源
徐永航
杜彦斌
李永亮
LIANG Qiang;XU Binyuan;XU Yonghang;DU Yanbin;LI Yongliang(School of Mechanic Engineering,Chongqing Technology and Business University,Chongqing 400067,China;Chongqing Key Laboratory of Green Design and Manufacturing of Intelligent Equipment,Chongqing Technology and Business University,Chongqing 400067,China)
出处
《表面技术》
北大核心
2025年第5期217-232,275,共17页
Surface Technology
基金
重庆市自然科学基金面上项目(cstc2020jcyj-msxmX0276)
重庆市高校创新研究群体资助项目(CXQT21024)。
关键词
激光表面硬化
随机森林算法
有限元模型
平均淬透深度
中心复合试验
多目标遗传算法
laser surface hardening
random forest algorithm
finite element model
average hardening depth
central composite test
multi-objective genetic algorithm
