摘要
针对激光熔覆过程中熔覆层深度无法精确控制问题,提出了基于海洋捕食者(Marine Predators Algorithm,MPA)优化的误差反向传播算法(Error Back Propagation,BP)单道激光熔覆熔深预测模型,以激光功率、扫描速度和送粉速率作为自变量,熔深作为因变量对模型进行评估。通过将该模型结果与PSO-BP、SOA-BP和SSA-BP神经网络的试验结果进行对比,发现MPA-BP预测模型的平均绝对误差为7.414%,拟合优度为0.964,相关数据的试验结果均优于其他模型,表明基于MPA优化的BP神经网络对熔深预测具有更好的稳定性和预测精度。
Aiming at the difficulty to accurately control the depth of cladding layer in laser cladding process,aprediction model of single channel laser cladding depth was proposed by the Error Back Propagation(BP)based on Marine Predator Algorithm(MPA)optimization.Laser power,scanning speed and powder delivery rate were taken as independent variables and penetration depth as dependent variables to evaluate the prediction model.By comparing the results of model with that of PSO-BP,SOA-BP and SSA-BP neural networks,the average absolute percentage error of MPA-BP prediction model is defined as 7.414%,and the goodness of fit is 0.964.The experimental results of relevant data are superior to those of other models,indicating better stability and prediction accuracy of BP neural network based on MPA optimization for penetration prediction.
作者
崔英浩
王梦乐
郭士锐
崔陆军
郑博
李晓磊
陈永骞
刘嘉霖
Cui Yinghao;Wang Mengle;Guo Shirui;Cui Lujun;Zheng Bo;Li Xiaolei;Chen Yongqian;Liu Jialin(School of Mechatronics Engineering,Zhengzhou Key Laboratory of Laser Additive Manufacturing Technology,Zhongyuan University of Technology)
出处
《特种铸造及有色合金》
CAS
北大核心
2023年第10期1425-1430,共6页
Special Casting & Nonferrous Alloys
基金
河南省水下智能装备重点实验室开放基金资助项目(YZC-2206-B0030-01-060)
河南省自然科学基金资助项目(202300410514)
中原工学院基本科研业务费专项资金资助项目(K2019QN006)
中原工学院青年骨干教师资助项目(2021XQG07)
中原工学院科研团队发展项目资助项目(K2021TD002)
河南省高等学校重点科研项目资助项目(21A460036)。
关键词
激光熔覆
海洋捕食者优化算法
预测模型
神经网络
Laser Cladding
Marine Predation Optimization Algorithm
Prediction Models
Neural Networks
