遗传算法优化BP神经网络的粒径大小研究

Study of Particle Size Based on BP Neural Network Optimized by Genetic Algorithm

为了提高湿法制粒所得药物片剂的生产质量,必须严格控制湿法制粒机出料口颗粒的平均粒度。该文根据系统的输入输出关系构建一合理的BP神经网络,用来拟合平均粒度大小与搅拌桨转速、切割刀转速、搅拌时间、切割时间非线性函数关系。此外,为避免BP神经网络由于初始权值、阈值的随机选取而导致的易陷入局部极值点而得不到全局最优解的缺点,利用遗传算法优化构建好的BP神经网络,从而实现对平均粒径大小与其影响因素的非线性拟合,实现对粒径的良好控制,提高片剂生产的质量。

In order to improve the quality of tablets produced by wet granulation,it is necessary to strictly control the average particle size of the granules in the outlet of the wet granulator. In this paper,we construct a reasonable BP neural network according to the input-output relationship of the system,which is used to fit the relationship between the average particle size and the mixing speed,cutting speed,stirring time and cutting time. In addition,in order to avoid the drawback that the BP neural network can not get into the global optimal solution due to the random selection of the initial weights and thresholds,we use the genetic algorithm to optimize the BP neural network. The non-linear fitting of the average particle size and its influencing factors can achieve good control of the particle size and improve the quality of tablet production.