基于机器学习的钢铁轧制过程性能预测

Property prediction of steel rolling process based on machine learning

为了实现快速的热轧工艺优化设计,基于工业数据的钢铁材料性能预测引起了研究者的极大关注,对利用机器学习进行钢铁材料轧制过程性能预测的研究进展进行了梳理。首先介绍了钢铁材料轧制过程性能预测常用的主流机器学习算法,其中包括人工神经网络、模糊神经网络、支持向量机、随机森林、智能优化算法等。其次,分别对钢铁材料轧制过程性能预测建模方法研究进展和模型应用情况进行了综述。最后,对钢铁轧制过程性能预测研究进行了展望,指出了数据质量的改善、小样本数据建模、建模数据加密、模型可解释性研究、钢铁材料组织预测和利用模型进行有效的工艺优化设计等可能发展方向。

In order to achieve rapid optimization design of hot rolling process,the property prediction of steel based on industrial data has attracted great attention of researchers.The research progress of steel rolling process property prediction using machine learning was reviewed.Firstly,the main machine learning algorithms commonly used in steel rolling process property prediction were introduced,including artificial neural network,fuzzy neural network,support vector machine,random forest,intelligent optimization algorithm and so on.Secondly,the research progress and the applications of steel rolling process property prediction model were summarized respectively.Finally,the prospect of the research on the property prediction of steel rolling process was presented,and the possible development directions were pointed out,such as the improvement of data quality,the modeling of small sample data,the encryption of modeling data,the research of model interpretability,the prediction of steel microstructure and the effective process optimization design by using the model.