中文摘要：

针对基本BP网络收敛速度慢、易陷于局部极小的不足，对基本BP网络的激活函数、动量项、学习率进行了较为全面的改进，并采用遗传算法进一步优化改进后的BP网络。分别采用基本BP网络、改进型BP网络、嵌入遗传优化的改进型BP网络建立了制品质量指标的预测模型，预测模型可根据输入的注塑工艺参数预测制品的质量指标。结果表明，嵌入遗传优化的改进型BP网络学习效率明显优于其他算法。

英文摘要：

In view of basic BP network lower convergence rate and easily to sink into the partial minimum insufficiency, this article has made more comprehensive improvement on the basic BP network＇s activation function, the momentum item, the study rate and so on, and adopted genetic algorithm to optimize the improved BP network. After the basic BP network, the improved BP network were used separately, based on built-in genetic algorithm optimization a forecasting model of part qualitative indexes was established. The model can forecast part qualitative indexes based on the modeling process parameter. The results indicate that the study efficiency of the genetic neural network surpasses other algorithms obviously.