基于置信度和自学习校正的热轧轧制节奏计算

针对目前国内大多数热连轧厂传统在线轧制节奏计算模型控制精度偏低的问题,在宝钢宁波钢铁热连轧厂对传统轧制节奏计算模型进行了分析与研究,提出一种基于置信度和自学习校正的热连轧轧制节奏计算模型改进算法.经过现场大量实测数据验证,采用新算法的轧制节奏计算模型的预报精度较传统模型算法有了大幅度的提高,偏差在±5s范围内的频率达到97.94%,大致成正态分布.新算法在预报带钢在线运行时间上优于传统模型算法.

Computation of Hot Tandem Rolling Rhythm Based on Confidence Level and Self-Learning Calibration

The control accuracy of conventional online rolling rhythm computation model is low in most hot tandem rolling plants in China at present. The problem was therefore investigated in the Baosteel Ningbo Hot Rolling Mill and, as a result, an improved algorithm of the rolling rhythm computation based on confidence level and self-learning calibration was proposed for hot tandem rolled strip. Verified by lots of in-situ testing data which were compared with that by the algorithm of conventional computation model, the prediction precision of the rolling rhythm computation model has greatly been improved, i.e., the error frequency is up to 97.94% within ±5 s and presents a normal distribution on the whole. In addition, the predicted time required for the whole rolling process of strip by the improved algorithm is superior to that by the conventional computation one.