激光熔凝参数BP神经网络的反求

发布时间：2018-01-20 16:37

本文关键词： 激光熔凝参数反求 BP神经网络热疲劳　出处：《哈尔滨理工大学学报》2017年03期 　论文类型：期刊论文

【摘要】：针对激光熔凝参数与熔凝单元体横截面尺寸之间呈现高度非线性映射关系,提出运用BP神经网络反求激光熔凝参数算法。利用BP网络算法建立激光熔凝参数反求模型,经过多次训练,模型预测误差缩小到3%以内。通过反求参数对DIEVAR模具钢做熔凝处理,结果表明,熔凝单元体横截面尺寸与期望值吻合误差为1.33%,能够满足期望精度。通过对比研究经反求参数做熔凝处理和未经处理DIEVAR材料的抗热疲劳性能,观察分析热疲劳裂纹在材料表面萌生及扩展形态,结果显示,经反求参数熔凝处理过的DIEVAR模具钢其抗热疲劳性能大幅度提高,熔凝单元体对裂纹阻断有很好的效果。
[Abstract]:There is a highly nonlinear mapping relationship between the laser melting parameters and the cross section size of the melting unit. In this paper, BP neural network is used to reverse the parameters of laser melting, and BP neural network is used to establish the model of parameters inversion of laser melting, which has been trained for many times. The prediction error of the model is reduced to less than 3%. The result shows that the cross section size of the DIEVAR die steel is consistent with the expected value and the error is 1.33%. By comparing the thermal fatigue resistance of DIEVAR materials with reverse parameters and untreated DIEVAR, the initiation and propagation morphology of thermal fatigue cracks on the surface of the materials were observed and analyzed. The results show that the thermal fatigue resistance of DIEVAR die steel treated by reverse parameter melting treatment is greatly improved, and the melting unit has a good effect on crack blocking.
【作者单位】：哈尔滨理工大学材料科学与工程学院;浙江大学宁波理工学院;江苏彤明车灯有限公司;
【基金】：国家自然科学基金(51275468) 浙江省自然科学基金(LY17E050013)
【分类号】：TN249;TP183
【正文快照】： 0引言通过激光快速熔凝可以细化材料组织,提高材料硬度、耐磨性、抗腐蚀性和抗热疲劳性能等[1-4]。由于其熔凝层薄,热作用区小,现已广泛应用于模具、汽车零部件等表面强化。激光熔凝强化时,激光对材料直接产生作用只有表层不到0.1 mm的深度,即金属表面受热熔化后,热量以热传导

【相似文献】