北京印刷学院学报
北京印刷學院學報
북경인쇄학원학보
JOURNAL OF BEIJING INSTITUTE OF GRAPHIC COMMUNICATION
2015年
4期
59-61
,共3页
神经网络%主成分分析法%氯丁橡胶%门尼黏度%预测
神經網絡%主成分分析法%氯丁橡膠%門尼黏度%預測
신경망락%주성분분석법%록정상효%문니점도%예측
neural networks%principle component analysis%chloroprene rubber%mooney viscosity%prediction
为了准确预测氯丁橡胶门尼黏度,采用氯丁橡胶生产工艺机理及数据分析结果相结合的方法,选取影响门尼黏度的主要因素,提出了一种基于 PCA-BP 神经网络的氯丁橡胶门尼黏度的软测量预测方法,建立了双隐含层的神经预测网络模型。通过训练,确定网络结构为6-5-13-1。仿真显示,最大相对误差为4.1%,平均相对误差为1.8%,最大绝对误差为4.389,误差在允许的范围内,满足生产要求。
為瞭準確預測氯丁橡膠門尼黏度,採用氯丁橡膠生產工藝機理及數據分析結果相結閤的方法,選取影響門尼黏度的主要因素,提齣瞭一種基于 PCA-BP 神經網絡的氯丁橡膠門尼黏度的軟測量預測方法,建立瞭雙隱含層的神經預測網絡模型。通過訓練,確定網絡結構為6-5-13-1。倣真顯示,最大相對誤差為4.1%,平均相對誤差為1.8%,最大絕對誤差為4.389,誤差在允許的範圍內,滿足生產要求。
위료준학예측록정상효문니점도,채용록정상효생산공예궤리급수거분석결과상결합적방법,선취영향문니점도적주요인소,제출료일충기우 PCA-BP 신경망락적록정상효문니점도적연측량예측방법,건립료쌍은함층적신경예측망락모형。통과훈련,학정망락결구위6-5-13-1。방진현시,최대상대오차위4.1%,평균상대오차위1.8%,최대절대오차위4.389,오차재윤허적범위내,만족생산요구。
In order to predict chloroprene rubber Mooney viscosity exactly,method of finding the main factors influencing Mooney viscosity is found combining production process mechanism of chloroprene rubber with the results of data analysis. A method of prediction of chloroprene rubber Mooney viscosity based on PCA and BP neural network is proposed and a double-hidden layer model of ANN is built. The network structure is proved to be 6-5-13-1 by training. The simulation shows that the maximum relative error is 4. 1%,the average relative error is 1. 8% and the maximum absolute error is 4. 389 which is in the permission range,and it can be used for conducting production.