基于遗传神经网络的钢管壁厚预测模型的研究

为解决钢管热轧过程荒管壁厚难以计算的问题,首先利用遗传算法优化了BP神经网络,在这种遗传神经网络算法的基础上建立了钢管轧制前毛管温度、长度、外径、轧辊转速、芯棒直径五项工艺参数与钢管轧制后荒管壁厚之间的数学模型.经过测试,基于遗传神经网络的钢管壁厚预测模型的壁厚预测误差远小于常规壁厚公式的计算误差,为设计更合理的设置毛管参数提供了科学的依据,对钢管热轧工艺水平的提高的具有重要意义.

Prediction model of pipe wall thickness based on genetic neural network

To solve the shortage of steel during hot rolling pipe wall thickness is difficult to calculate the problem, the first to use genetic algorithms optimize BP neural network, in which genetic neural network algorithm based on a tube -roiling before the capillary temperature, length, diameter, roll rotational speed, mandrel diameter of five process parameters and tube - roll- ing pipe wall thickness after the shortage between the mathematical model. Tested, based on genetic neural network prediction model for wall thickness of pipe wall thickness prediction error is much smaller than conventional wall thickness formula for the calculation error, a more reasonable setting for the design of capillary parameters provide a scientific basis for the level of hot - rolling steel pipe improvement of great significance.