摘要
为了解决在没有数控程序的情况下快速准确的预测飞机结构件的数控加工工时,提出了一种基于特征与遗传神经网络的数控加工工时预测方法。提炼各类特征的加工工时影响因素并建立特征样本库,为每种加工特征构建BP神经网络;针对BP神经网络极易陷入局部极小值、收敛速度慢、网络参数难以确定等问题,结合遗传算法优化BP神经网络。建立了5种神经网络结构,通过调用相应的网络预测每一类加工特征的加工工时,进而形成每一工步的加工工时,累加零件所有工步的加工工时得到零件整体的加工工时。应用该方法预测零件整体的加工工时误差在5%以内。
In order to rapidly and accurately forecast man-hour of aircraft structure parts without NC programs, a method for forecasting man-hour based on the features and genetic algorithm are proposed in this paper. A database is established according to the influencing factors refined here which determine the man-hour of every feature. This method combines back propagation （BP） neural network with genetic algorithm to deal with the defects of the steepest descent in slowly converging, easily immerging in partial minimum frequently and difficultly determining the optimal parameters. A total of 5 structures of neural networks are established with regard to typical processing technology of aircraft structure parts. Man-hour of every machining feature is forecasted by using corresponding structure, the total of which form the whole man-hour of the parts. The simulation result shows high practical value of this method with error below 5%.
出处
《机械科学与技术》
CSCD
北大核心
2014年第7期1111-1116,共6页
Mechanical Science and Technology for Aerospace Engineering
基金
国家科技重大专项项目(2012ZX04010041)资助
关键词
飞机结构件
加工特征
工时预测
BP神经网络
遗传算法
aircraft structural parts
aircraft
backpropagation
computer simulation
database systems
errors
flowcharting
forecasting
genetic algorithms
machining features
machining
man-hour forecasting
MATLAB
neural networks
schematic diagrams
