基于GMM-KNN-LSTM的烧结矿化学指标预测

Prediction of Sinter Chemical Indexes Based on GMM KNN-LSTM

针对烧结矿化学指标检测频率低导致无标签样本无法被机器学习利用的问题,提出了一种充分利用样本中有用信息的烧结矿化学指标预测模型.首先,结合高斯混合模型(GMM)和K-近邻(KNN)算法,将无标签样本转化为有标签样本,然后与长短期记忆(LSTM)单元相结合,用于预测烧结矿的总铁质量分数、FeO质量分数和碱度3个化学指标.通过与反向传播神经网络(BPNN)、循环神经网络(RNN)和LSTM三种模型对比,结果表明所建模型具有较低的预测误差.总铁质量分数和FeO质量分数的预测命中率在允许误差±0.5%内时分别达到98.73%和95.33%,碱度的预测命中率在允许误差±0.05内为98.13%,展现了较高的预测精度.

Aiming at the problem that unlabeled samples cannot be utilized by machine learning due to the low detection frequency of sinter chemical indexes,a prediction model for sinter chemical indexes that makes full use of the useful information in the samples is proposed.Firstly,the unlabeled samples are transformed into labeled samples by combining Gaussian mixture model(GMM)and K-nearest neighbor(KNN)algorithm,and then combined with long short-term memory(LSTM)unit for predicting three chemical indexes,namely,total Fe mass fraction,FeO mass fraction,and alkalinity of sinter.By comparing with the three models of back propagation neural network(BPNN),recurrent neural network(RNN),and LSTM,the results show that the proposed model has a low prediction error.The prediction hit rates of total Fe mass fraction and FeO mass fraction reach 98.73%and 95.33%,respectively within the allowable error of±0.5%,and the prediction hit rate of alkalinity is 98.13%within the allowable error of±0.05,demonstrating high prediction accuracy.