摘要
为提高液压缸零件工时定额预测的准确性及高效性,提出一种基于加工特征参数的DE_kmeans预测模型。首先根据工时影响因素提炼出历史及待预测加工特征参数;采用改进的DE_kmeans算法对历史加工特征参数进行聚类成组;对每个聚类组分别建立BP神经网络预测模型并基于历史加工特征参数进行训练;针对待预测加工特征参数,按照标准化欧式距离最小的原则划分至特定聚类组及预测模型;用该模型对待预测零件工时进行预测。通过测试实例验证该方法的预测误差控制在10%以内,证明该方法的可行性及有效性。
To improve the forecasting accuracy and efficiency of man-hour quota with hydraulic cylinder parts, a model based on DE_kmeans algorithm was proposed by analyzing processing characteristics. Firstly, the historical and unsettled processing characteristics were figured out by man-hour influencing factors. Secondly, clustering groups were generated by applying improved DE_kmeans algorithm on historical processing characteristics. Thirdly, forecasting models corresponding to different clustering groups were set up respectively by means of BP neural network with historical processing characteristics trained. Fourthly, unsettled processing characteristics were divided into particular clustering group and model by principle of standardized euclidean distance minimization. Finally, man-hour quota forecasting data were obtained based on this particular model for estimation. Algorithm proposed in this paper was tested to verify its feasibility and effectiveness by the error within 10%.
作者
潘彩霞
陆宝春
张均利
PAN Cai-xia;LU Bao-chun;ZHANG Jun-li(School of Mechanical Engineering Nanjing University of Science and Technology, Jiangsu Nanjing 210094, China)
出处
《机械设计与制造》
北大核心
2019年第4期162-165,共4页
Machinery Design & Manufacture
