基于改进的AdaBoost.RS算法的烧结终点预报分析

烧结终点的稳定控制是提高烧结机利用效率及烧结矿产量和质量的前提，因此获得准确的烧结终点位置是优化烧结过程的基础。通过分析烧结过程参数对烧结终点位置的影响，提出一种适用于烧结终点预测的集成算法。在AdaBoost.RS算法的基础上，自适应调整松弛变量的阈值，以极限学习机为弱学习器建立烧结终点位置预报集成算法模型。以宝钢烧结面积为495 m2的烧结机为例，利用实际生产数据进行模型检验。结果表明，当绝对误差小于1.6 m时，模型的预报结果命中率为97.4%，均方根误差为0.58，预报值序列与实际目标值序列的相关系数为0.78。对各影响因素定量分析结果表明，影响烧结终点位置的前三因素依次为料层厚度、台车速度与配水量。

Prediction and analysis of burning though point base on #br# modified AdaBoost.RS algorithm

The accurate control of burning though point is the premise to improve the utilization efficiency of the sintering machine and the output and quality of sintering agglomerate. Therefore, obtaining the accurate burning though point position is the basis of optimizing sintering process. By analyzing the effect of the sintering process parameters on the position of burning though point, an integrated algorithm for predicting burning though point was proposed. Base on the AdaBoost.RS algorithm, the threshold of the slack variable was self adjusted. An integrated predictive model for predicting burning though point was established by using extreme learning machine as a weak learner. The proposed method was verified by using the actual production data derived from a 495 m2 sintering machine in Baosteel. The experimental results showed that when the absolute error was less than 1.6 m, the hit rate of the model was 97.4%, the root mean square error was 0.58, and the correlation coefficient between the predicted value series and the actual target value series was 0.78. Quantitative analysis of the influencing factors showed that the front three factors affecting the burning through point were layer thickness, trolley speed and water distribution.