摘要
为了优化往复泵柱塞在激光加工表面仿生织构前期所开展的激光加工参数最优组合筛选方法,开展了多组柱塞表面激光加工仿生织构试验,基于试验数据建立并成功训练出了激光加工参数映射织构体积的BP人工神经网络预测模型。结果表明,激光功率、扫描速度、烧结次数对织构体积的影响具有互相独立性和等效关联性;隐含层含有4个神经元的三层BP神经网络模型的平均预测精度约为92%,可满足实际应用要求。此研究结果在实现激光加工参数到织构体积正向预测的基础上,对仿生织构技术在压裂泵柱塞表面上的工业化批量应用有重要意义。
In order to optimize the optimal combination screening method of laser processing parameters in the early stage of laser processing surface bionic texture of reciprocating pump plunger,a laser engraving texture experiment on the surface of the plunger was carried out.Based on the experimental data,a BP artificial neural network prediction model between laser parameters and texture volume was established and successfully trained.The results show that the effects of laser power,travel speed,and scanning pass on the texture volume are mutually independent and equivalently related.Also,the average prediction accuracy of the three-layer BP neural network model with 4 neurons in the hidden layer is about 92%.Based on the forward prediction of laser parameters to texture volume,this research is of great significance for the industrial batch application of bionic texture technology on the surface of fracturing pump plunger.
作者
陈林燕
王腾
王国荣
曾兴昌
Chen Linyan;Wang Teng;Wang Guorong;Zeng Xingchang(School of Engineering,Southwest Petroleum University,Nanchong,Sichuan 637001,China;School of Mechatronic Engineering,Southwest Petroleum University,Chengdu,Sichuan 610500,China;Energy Equipment Institute,Southwest Petroleum University,Chengdu,Sichuan 610500,China;CNPC National Oil&Gas Drilling Equipment Engineering Technology Research Center,Baoji,Shaanxi 721002,China)
出处
《应用激光》
CSCD
北大核心
2022年第11期57-63,共7页
Applied Laser
基金
国家自然科学基金资助项目(51775463)
南充市校科技战略合作项目(18SXHZ0049)。
关键词
激光技术
织构化柱塞
激光加工
神经网络
laser technique
textured plunger
laser engraving
neural network
