基于正交实验和数据驱动的喷丸表面完整性参数预测

Prediction of Surface Integrity Parameters of Shot Peening Based on Orthogonal Experiment and Data-driven

目的探究喷丸工艺参数对18CrNiMo7-6滚子表面完整性的影响规律,得到喷丸工艺参数与表面完整性的映射关系,提高喷丸工艺的质量与效率。方法运用Python语言对Abaqus进行二次开发,建立喷丸仿真的随机多弹丸模型并进行了试验验证。设计正交实验研究喷射角度、喷射速度、弹丸直径、覆盖率及弹丸类型对残余应力与表面粗糙度的影响规律,并用随机森林算法得到各个工艺参数对喷丸综合效果的重要度值。以喷射角度、喷射速度、弹丸直径、覆盖率、弹丸类型、距表面深度为输入,残余应力和表面粗糙度为输出,建立基于神经网络的喷丸表面完整性参数预测模型。结果通过正交实验分析得到弹丸直径和喷射速度对表面粗糙度有显著影响。各个喷丸工艺参数对18CrNiMo7-6滚子的喷丸综合效果的重要度依次为:喷射角度0.249,喷射速度0.224,弹丸类型0.193,覆盖率0.173,弹丸直径0.161。在各个工艺参数范围内,较优的工艺参数组合为:喷射角度90°,喷射速度80 m/s,弹丸直径0.7 mm,覆盖率300%,弹丸材料为铸钢丸。基于神经网络的喷丸表面完整性参数预测模型的平均相对误差低于7%。结论基于神经网络的喷丸表面完整性参数预测模型可以较准确地表示喷丸工艺参数与表面完整性参数之间的映射关系,能够为喷丸工艺提供相关参考。

This paper aims to study the influence of process parameters of shot peening on the surface integrity of 18CrNiMo7-6 roller and acquire the mapping relation between process parameters and surface integrity,so as to improve the quality and efficiency of shot peening process.During this,Python language was used for the secondary development of Abaqus to establish the random multi-shots model of shot peening simulation and the experimental verification was carried out.Orthogonal experiment was designed to study the effect laws of impact angle,impact velocity,shot diameter,coverage and shot type on residual stress and surface roughness,and the importance value of each process parameter on the comprehensive effect of shot peening was obtained by the random forest algorithm.With impact angle,impact velocity,shot diameter,coverage,shot type and surface depth as input values and residual stress and surface roughness as output values,a prediction model of shot peening surface integrity based on neural network was established.Through the orthogonal test,it is found that the shot diameter and impact velocity have a significant influence on the surface roughness.The importance of each shot peening process parameter to the comprehensive shot peening effect of 18CrNiMo7-6 roller is impact angle(0.249),impact velocity(0.224),shot type(0.193),coverage(0.173)and shot diameter(0.161).The optimal combination of process parameters within the range of each process parameter is that the impact angle is 90°,the impact velocity is 80 m/s,the shot diameter is 0.7 mm,the coverage is 300%,and the shot material is cast steel shot.The average relative error of the shot peening surface integrity prediction model based on neural network is less than 7%.Therefore,it is concluded that the shot peening surface integrity prediction model based on neural network can accurately represent the mapping relation between the shot peening process parameters and surface integrity parameters,thus providing relevant reference for shot peening process.