基于即时学习算法的铸坯质量预测方法

传统铸坯质量预测大多基于全局建模方法,存在自适应能力较差、预测精度不稳定等不足。为此,提出了一种基于即时学习算法的铸坯质量预测方法,主要特征为构建基于即时学习算法的局部模型,以取代传统全局模型,实现建模和预测同时在线进行,更适应于场景复杂、工况多变的连铸生产过程;根据连铸生产数据的时变性特点,在相似度计算中引入时间权重因子,强化样本数据与待测数据的相关性,更有利于提高铸坯质量预测模型精度。以国内某钢厂65号高碳钢铸坯三角区裂纹为例,具体说明即时学习算法在铸坯质量预测局部模型构建中的应用,并与铸坯质量预测全局模型的预测结果进行对比分析及评价。结果表明,基于即时学习算法的局部模型在评价指标上均优于全局模型,全局模型的预测准确率为65%,基于即时学习算法的局部模型准确率提升至90%,进一步阐明基于即时学习算法的局部模型用于铸坯质量预测的有效性。

Predictive method of casting slab quality based on just-in-time learning algorithm

Traditional models for the prediction of casting slab quality were mostly built using the global modeling method, which had poor self-adaptive ability and unsteady prediction accuracy. In this study, a new method based on the just-in-time learning algorithm was proposed for the prediction of casting slab quality. The main feathers of the method included the local model based on the just-in-time learning algorithm was built to replace the traditional global model, and the just-in-time modeling method make the prediction model more adapted to the continuous casting process with complex production scenarios and various working conditions. According to the time-varying characteristics of continuous casting production data, the time weighting factor was introduced into the similarity calculation to strengthen the correlation between the sample data and the data to be predicted, which was more beneficial to increasing the model accuracy for the prediction of casting slab quality. Taking the triangular crack of No.65 high-carbon steel slab in a steel plant as an example, the application of just-in-time learning algorithm in the construction of local model for casting slab quality prediction is illustrated, and the prediction results are compared with the global model for casting slab quality prediction. The results showed that the performance of the local model was better than that of the global model based on the assessment indexes. The prediction accuracy of the global model was 65%, while that of the local model was increased to 90%. The effectiveness of the local model for the prediction of casting slab quality was confirmed by comparing the prediction results.