BP改进算法神经网络的保护渣性能预测模型

采用混料回归设计原理设计试验保护渣组成。以试验为基础，针对常用BP算法的不足采用动量因子与自适应学习速率相结合的BP改进算法建立神经网络保护渣性能预测模型。研究结果表明该模型预测精度高，适用组元多、成分变化范围大；对保护渣的性能预测取得了较好的效果，能为保护渣设计提供理论指导。

Predictive model of mould flux properties based on improved BP algorithm of neural network

Compositions of the mould flux are designed according to the principle of regression design on the blended mix. In light of deficiency of the normal BP algorithm, a neural network predictive model for the mould flux properties is established on the basis of the improved BP algorithm in combination of momentum factor and self-adaptive learning rate. Experimental results show that this model is highly accurate and suitable for the mould flux with multi-components and the wide range of compositions. Better results have been achieved in prediction of the mould flux properties by this new model, thus providing a solid theoretical basis for mould flux design.