基于主成分分析的BP神经网络内螺纹冷挤压成形质量预测  

Forming Quality Forecast for Internal Threads Formed by Cold Extrusion Based on Principal Component Analysis and Neural Networks

作　　者：张敏[1] 黎向锋[1] 左敦稳[1] 缪宏[1] 

机构地区：[1]南京航空航天大学,南京210016

出　　处：《中国机械工程》2012年第1期51-54,共4页China Mechanical Engineering

基　　金：空军装备部"十一五"预研项目

摘　　要：根据冷挤压内螺纹成形中径、螺距、牙型半角和牙高率等来综合评定内螺纹的成形质量等级,并基于BP神经网络对其进行预测。在BP神经网络预测模型数据前处理过程中,采用主成分分析方法以提取影响内螺纹冷挤压成形质量的主要因素,消除各影响因素之间的线性相关性。试验结果表明,与传统的BP神经网络相比,采用该方法的BP神经网络模型,简化了网络结构,提高了收敛速度及预测精度,能准确实现内螺纹成形质量等级的预测,从而为内螺纹质量的检测提供了一条新途径。Forming quality grade of internal threads formed by cold extrusion was rated synthetically by BP neural network based on pitch diameter, thread pitch, half of thread angle and threads height ratio. For eliminating linear relevance in inter-influencing factors while the data pre-processing, major factors that affected the forming quality of internal threads formed by cold extrusion were extracted by principal component analysis. The experimental results show that the neural networks input by the processed data by this method become simple, with improved convergence rate and forecast accuracy. This method realizes the forecast for quality grade of internal threads formed by cold extrusion precisely. Also it provides a new solution for detection of internal thread quality.

关 键 词：内螺纹 成形质量预测 主成分分析 神经网络 

分 类 号：TG376[金属学及工艺—金属压力加工] O235[理学—运筹学与控制论]