基于遗传神经网络的工时定额系统_郭超.pdfVIP

第 27 卷第 8期 计算机应用与软件 V o l127 No. 8 2010 年 8月 Com puter A pplications and Sofwt are A ug. 2010 郭 超 周丹晨 ( 62 1900 为提高工时定额的计算效率和准确度, 通过对工时定额表的特点分析, 设计开 了基于神经网络的工时定额系统, 并针 对 BP 神经网络存在的易陷入局部最小值不易收敛全局有哪些信誉好的足球投注网站能力差等缺点, 将遗传算法集成到系统之中介绍了系统的主要功 能和工作流程, 详细论述了隐层结点数目的计算样本的自动划分遗传算法优化脱离M ATLA B 环境神经网络模型的保存等关键 技术实际应用证明系统具有计算精度高可以独立于MATLAB 运行通用性强等特点 工时定额 BP 神经网络 标准遗传算法 MATCOM MAN-HOUR QUOTA SYSTEM BASED ON GENETIC NEURAL NETW ORK Guo Chao Zhou Danchen ( Institute of M echanicalManuf acture Technology, Ch ina Academy of Eng ineering P hysics, M ianyang 621900, ichuan, China) Abstract T o mi prove com putation e ff iciency and accuracy o f man-hour quo ta, a m an-hour quota system based on neura l newt ork w as ex- p lored through the ana ly sis of the characteristics o f the m an-hour quota table. For the BP neura l ne wt ork has d isadvantages like getting into lo- ca lm in mi um po int easily, low convergence speed, w eak g loba l search ability and so on, gene tic algorithm w as integ rated into the system. T he pr mi ary functions and w ork flow of the system w ere introduced, and the key techno log ies w ere exp la ined, inc lud ing ca lcu la tion of hide layer node, autom atic sam ple div ision, g enetic algorithm s optmi ization, breaking aw ay from the env ironm ent of MATLAB, the sto rage o f neura l net- w ork m ode ls. P ractica l applica tion show s that the system has h igh computational accuracy and un ive rsality, and cou ld run w ithout the suppo rt o f M ATLAB. K eywor s M an-hour quo ta BP neural netw o rk Standard genetic a lgo rithm M ATCOM [ 8]