微计算机信息
微計算機信息
미계산궤신식
CONTROL & AUTOMATION
2010年
5期
39-41
,共3页
轮速信号%BP神经网络模型%优化设计
輪速信號%BP神經網絡模型%優化設計
륜속신호%BP신경망락모형%우화설계
wheel speed signal%BP neural network model%optimizing design
本文将神经网络方法引入到轮速信号处理之中,以实际采集的噪声信号作为输入,以小波滤波信号作为标准的输出,设计一个3层BP网络,根据结构的优化设计新方法,建立相应的神经网络模型:以所采集教据中的一部分作为学习样本,对所建神经网络模型进行训练、仿真;以采集数据中的其余部分作为检验数据.仿真结果表明,该模型能以很小的误差逼近标准的滤波输出.
本文將神經網絡方法引入到輪速信號處理之中,以實際採集的譟聲信號作為輸入,以小波濾波信號作為標準的輸齣,設計一箇3層BP網絡,根據結構的優化設計新方法,建立相應的神經網絡模型:以所採集教據中的一部分作為學習樣本,對所建神經網絡模型進行訓練、倣真;以採集數據中的其餘部分作為檢驗數據.倣真結果錶明,該模型能以很小的誤差逼近標準的濾波輸齣.
본문장신경망락방법인입도륜속신호처리지중,이실제채집적조성신호작위수입,이소파려파신호작위표준적수출,설계일개3층BP망락,근거결구적우화설계신방법,건립상응적신경망락모형:이소채집교거중적일부분작위학습양본,대소건신경망락모형진행훈련、방진;이채집수거중적기여부분작위검험수거.방진결과표명,해모형능이흔소적오차핍근표준적려파수출.
The paper applies the neural network technique to the signal processing of automobile wheel speed, a BP neural network of three layer has been designed with the noise collected actually as input, the signal filterred by wavelet as output, so the corre-sponding neural network model is setted up according to a new method of optimizing design; then the neural network model is trained and simulated with a part of data which is collected as study sample, the other data as test data, The simulating results show that the model can approach normal filtering output with small error.