BP神经网络与广义RBF神经网络在产品寿命分布模型识别中的应用研究  

Application Study on BP Network and Generalized RBF Network in Estimating Distribution Model of Mechanical Products

作　　者：吴月明[1] 王益群[1] 李莉[1] 

机构地区：[1]燕山大学,秦皇岛066004

出　　处：《中国机械工程》2006年第20期2140-2144,共5页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(60474044);河北省自然科学基金资助项目(E2004000221)

摘　　要：通过分析BP神经网络和广义径向基神经网络(RBF神经网络)的基本结构和算法,建立了应用于机械产品的寿命分布类型智能识别的BP神经网络和RBF神经网络模型,通过仿真实验对比分析了BP神经网络和RBF神经网络识别效能。结果表明,RBF神经网络参数调整简单、训练时间短,且逼近能力与BP神经网络相当,但是推广能力较差;BP神经网络收敛缓慢、训练时间长,但推广能力较强,更适合分布参数变化范围大的寿命分布模型的识别。A method of estimating distribution model based on BP network and generalized RBF network was proposed, and a simulation example was presented. The results obtained from the simulation indicate that the function approximation capability of BP network and generalized RBF network is almost equal. RBF network has a simple way to train the weights, so its convergence is prior to that of BP network. In the other hand, BP network needs smaller space occupation and its generalization ability is better than that of RBF network when solving the same problem.

关 键 词：BP神经网络 径向基神经网络 可靠性 寿命分布模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]