【摘要】 采用变步长和惯性项调整权值系数及阔值的改进BP算法,建立了一润滑油调合BP神经网络系统。分别预测了内燃机稠化油100℃运动粘度(V100)和润滑油的配方组成。预测结果的误差分别为:内燃机稠化油V100的绝对误差(A.D.)在±0.1mm2/s范围内,相对误差(R.D.)在±1.5％范围内;润滑油调合各组分质量百分含量的绝对误差在±1.2％范围内,相对误差在±2.0％范围内。结果表明BP神经网络预测误差能满足试验要求,预测精度优于常规非线性回归方程(分别为:V100 A.D.在±0.5mm2/s范围内,R.D.在±4.0％范围内;组分含量A.D.在±3.0％范围内,R.D.在±4.0％范围内)。为润滑油调合和相关数据的计算提供了一种新方法。更多还原

【Abstract】 A BP neural network system using for the concoction of the lubricating oil was developed firstly in the paper, the BP algorithm was applied in the course of the development by adjusting weight coefficient and threshold on the basis of variable step and inertia item. This BP neural network had been used successfully to predict the viscosity (100t) of the diesel bodying oil and to optimize the formulation of lubricating oil, respectively. The deviations of the predicting results are as following. The absolute deviations (A. D.) of the viscosities (100t) of the diesel bodying oil are in the range of ±0.1mm2/s, and the relative deviations are in the range of(R. D.) ±1.5%. The A. D. of the percent weight of the components in the lubricating oil are in the range of ±1.2%, and the R. D. are in the range of ±2.0% . The results have proved that the BP neural network is a favorable non-modeled method and can be satisfied with the prediction. It also has the advantage with higher predicting accuracy than nonlinear regression (The A. D. of V100 in the range of ± 0.5mm2/s and the R. D. in the range of ±4.0% . The A. D. of the percent weight in the range of ±3.0% and the R. D. in the range of ±4.0%).更多还原