摘要
通过试验测定焊条熔敷金属的断后伸长率和冲击吸收功等力学性能,以多层前馈神经网络(BP)、径向基函数神经网络(RBF)、自适应模糊神经网络(AFNN)3种方法获得的单一预测模型的计算值作为输入参数,建立焊条熔敷金属力学性能神经网络非线性组合预测模型.利用42组试验样本对模型进行训练和验证.结果表明,断后伸长率和冲击吸收功的预测平均相对误差均在5%以内,满足实际生产要求.采用Matlab和visual C++混合编程技术开发了基于数据库的焊条性能智能预测软件系统,可以直接根据焊条原材料成分对焊条熔敷金属的断后伸长率、冲击吸收功等力学性能进行较为准确的预测,为焊条的质量预测与控制提供了一条简单有效的新途径.
The elongation after fracture and impact energy of the deposited metal by electrode were tested by experiments.A nonlinear combination neural network model to predict the deposited metal mechanical properties by electrode was built by taking predicted data acquired by such single models as BP,RBF and adaptive fuzzy neural network as input parameters.42 groups of experimental samples were used to train and verify the model.The results show that the average relative prediction errors of both elongation after fracture and impact energy are less than 5% and it satisfies the demands of practical production.An intelligent prediction system of electrode properties was developed by Matlab and visual C++ and it can correctly predict the elongation after fracture and impact energy of deposited metal by electrode according to the raw material components.It provides a new,simple and effective way to predict and control the electrode quality.
出处
《焊接学报》
EI
CAS
CSCD
北大核心
2011年第5期89-92,117,共4页
Transactions of The China Welding Institution
关键词
自适应模糊神经网络
熔敷金属
碳钢焊条
非线性组合预测
adaptive fuzzy neural network
deposited metal
carbon steel electrode
nonlinear combination prediction
