基于改进神经网络的机身镀层抗冲击性能预测  

Impact resistance prediction of fuselage coatings based on improved neural network

作　　者：邱有春[1] 赵立平 QIU Youchun;ZHAO Liping(Luzhou Vocational and Technical College,Luzhou 646000,China;Weifang Institute of Technology,Qingzhou 262500,China)

机构地区：[1]泸州职业技术学院,四川泸州646000 [2]潍坊理工学院,山东青州262500

出　　处：《兵器材料科学与工程》2022年第5期148-153,共6页Ordnance Material Science and Engineering

摘　　要：为准确预测机身镀层抗冲击性能,以40Cr钢为机材,用超声电沉积技术制备Ni-SiC纳米镀层,并进行真空热处理,用平头弹、尖头弹冲击Ni-SiC纳米镀层。用粒子群算法改进RBF神经网络,结合AdaBoost算法构建机身镀层抗冲击性能的预测模型。将Ni-SiC纳米镀层工艺参数及冲击速度作为模型输入,进行Ni-SiC纳米镀层的抗冲击性能预测。结果表明:改进RBF神经网络的最优网络结构为3-10-1,预测误差为0.055%~1.570%,预测精度高;在SiC纳米粉体为8 g/L、电流密度为3 A/dm^(2)、镀液温度为40℃条件下制备的Ni-SiC纳米镀层形貌最优;以不同速度冲击Ni-SiC纳米镀层,平头弹均未断裂,尖头弹不同程度断裂。In order to accurately predict the impact resistance of fuselage coating,the Ni-SiC nano coating was prepared by ultrasonic electrodeposition technology with 40Cr steel as the substrate,and then subjected to vacuum heat treatment.The NiSiC nano coating was impact by flat head bullets and pointed bullets.The RBF neural network was improved by particle swarm optimization algorithm,and the predictor model of the impact resistance of the fuselage coating was constructed by combining AdaBoost algorithm.The process parameters and impact velocity of Ni-SiC nano coating were used as model inputs to predict the impact resistance of Ni-SiC nano coating.The results show that the optimal network structure of the improved RBF neural network is 3-10-1,the prediction error is 0.055%-1.570%,and the prediction accuracy is high.The morphology of Ni-SiC nanocoating obtained under the conditions of SiC nano-powder 8 g/L,current density 3 A/dm^(2) and bath temperature 40℃is optimum.When the Ni-SiC nano coating is impacted at different speeds,the flat bullets is not broken,and the pointed bullet is broken to different degrees.

关 键 词：改进神经网络算法 机身镀层 抗冲击性能 超声电沉积技术 预测准确度 

分 类 号：TG174.4[金属学及工艺—金属表面处理]