摘要
激光切割金属板材过程中,工艺条件和表面质量之间存在较为复杂的对应关系;提出基于BP神经网络的激光切割质量控制模型,建立工艺条件与切割面粗糙度之间的关系模型,试验测量点取距切割下边缘1.5mm处表面粗糙度Ra;提出利用模拟退火算法提高多层神经网络的拟合精度,改善网络的收敛性能;切割试验样本设计拟采用星点设计法,用于提高神经网络训练样本的信息量和可靠性;经过实际切割不锈钢板材,验证上述方法具有一定的可靠性和应用价值。
In the process of laser cutting metal sheet,there was complex correspondence between technological parameter and surface quality.The control model of laser cutting quality was proposed based on BP neural networks,and relational model between technological parameter and surface roughness was established.The measure point for measuring surface roughness Ra is located at 1.5mm distance from lower cutting edge.By the use of the simulated annealing,the fitting accuracy of network and network convergence were improved.Design for cutting test sample adopted the center composite design for improving the amount of information and reliability of network training samples.After the practical cutting of stainless steel sheet,the result verified the reliability of the method and value of application.
出处
《机械设计与制造》
北大核心
2012年第6期85-87,共3页
Machinery Design & Manufacture
基金
南京工程学院科研基金资助项目(KXJ08141)
江苏六大人才高峰第六批资助项目-三菱可视化e-f@ctory自动化生产线的研究成果
关键词
激光切割
表面质量
神经网络
模拟退火
Laser Cutting
Surface Qualification
Neural Network
Simulated Annealing
