CAJ | 학술논문

针对传统BP算法存在的不足进行改进，采用共轭梯度法与Levenberg-Marquardt法对BP神经网络进行优化；通过实际数据进行预处理、建模分析，对比传统BP神经网络和经过优化后BP神经网络，证明了优化后的神经网络在油品污染与磨损的预测方面具有更好的泛化能力。
침대전통BP산법존재적불족진행개진，채용공액제도법여Levenberg-Marquardt법대BP신경망락진행우화；통과실제수거진행예처리、건모분석，대비전통BP신경망락화경과우화후BP신경망락，증명료우화후적신경망락재유품오염여마손적예측방면구유경호적범화능력。
Aimed for improvement of the shortcomings existed in the traditional BP algorithm,the BP neural network was opti-mized by using conjugate gradient method and Levenberg-Marquardt method.Through the actual data pre-processing,modeling and a-nalysis,the traditional BP neural network and optimized BP neural networks were compared.It is proved that the optimized neural net-work has better generalization ability in the aspect of oil pollution and prediction of the wear.