基于最优组合算法的烧结终点集成预测模型

Integrated model for predicting burning through point of sintering process based on optimal combination algorithm

针对铅锌烧结工艺过程 ,提出一种集成的烧结终点预测模型。首先利用具有自适应能力和逼近任意非线性函数的神经网络技术建立烧结终点预测模型 ;其次利用总结专家知识和操作者经验的模糊逻辑规则建立烧结终点预测模型 ;最后采用最优组合算法综合集成两种模型作为烧结终点的预测模型。实际预测结果表明 :利用集成方法建立的预测模型能够较为准确地反映实际生产过程 。

For the complicated Pb Zn sintering process, an approach that integrates two models to predicting the burning through point(BTP) was presented. The first method uses BP neural network and Levenberg Marquardt algorithm to predict BTP; the second method uses metallurgical and control expert knowledge and skilled operator experiences of the sintering process. Two results are combined by using optimal combination algorithm as general modeling results of the burning through point. Through the analysis of the industrial site modeling experiment, the results show that the integration of two model can show the real time production exactly.