基于极限学习机的板形建模与优化

Plate shape modeling and optimization based on extreme learning machine

以铝合金板带生产厂2100 mm轧机的4.1 mm厚铝薄板带轧制生产过程作为研究对象,综合分析各个因素对板形的影响规律,分别从平直度和截面形状两个方面建模和研究,采用平直度转换为与标准板形曲线的厚度差,建立了四辊轧制过程中铝薄板带板形预测模型,预测误差范围为-0.0223~0.0191 mm。为了进一步提高模型预测精度,采用极限学习机智能算法对预测模型进行修正,修正后的预测模型误差更小,为-0.0160~0.0141 mm。考虑了平直度和板厚分布的综合预测模型可以更好地反映板形,同时为高精度板带轧制过程中板形在线控制提供了理论基础。

Taking the rolling process of aluminum thin strip with the thickness of 4.1mm for the 2100 mm rolling mill in an aluminum alloy strip production plant as the research object,the influence of various factors on the strip shape was comprehensively analyzed.The flatness and section shape were modeled and studied respectively.The flatness was converted into the thickness difference with the standard strip shape curve,and the shape prediction model of aluminum strip in the 4-high mill rolling process was established.The prediction error range was-0.0223~0.0191 mm.In order to further improve the prediction accuracy of the model,the extreme learning machine intelligent algorithm was used to modify the prediction model,and the error of the modified prediction model was smaller,reaching-0.0160~0.0141 mm.The comprehensive prediction model considering flatness and thickness distribution can better reflect the shape,and provide a theoretical basis for on-line shape control in high-precision rolling process of strips.