Abstract: The iron-oxide scale total thickness and FeO layer thick percentage of 72A high carbon steel (0.67%-0.73% C) coils have markedly influence on coil stripping performance. With using artificial neural networks and mathematical method the BP network model is established to realize the complex non-linear mapping relationship between the cooling schemes during oxidation process of high carbon steel wire and the formed FeO layer thickness percentage after oxidation. Compared between measured parameters and network simulation results, it is obtained that the training accuracy of BP network is very high and has better generalization ability to quite well response the effect of each factor on FeO layer thickness percentage. The production application results show that based on BP network simulation results to improve cooling process, suitable decreasing wire laying-off temperature and increasing cooling speed at 850-720℃，the FeO layer thickness percentage of scale decreases in order to improve the mechanical peeling performance.

Key words: High Carbon Steel 72A Coil, Peeling Performance, Artificial Neural Network, Cooling Scheme, FeO Layer Thickness Percentage

摘要： 72A高碳钢（0.67%~0.73%C）盘条的氧化铁皮总量和氧化铁皮中FeO层厚百分比显著影响盘条的剥离性能。利用人工神经网络和数理方法,建立BP网络模型,实现了高碳钢线材氧化过程中的冷却制度与氧化后生成的FeO层厚百分比之间的复杂的非线性映射关系。将实测的参数与网络模拟结果进行比较得出,建立的BP网络训练精度非常高,泛化能力强,能很好的反应各个因素对FeO层厚百分比的影响。生产应用结果表明,根据BP网络模拟结果改进冷却工艺,适当降低吐丝温度,提高850~720℃区间冷却速度,使氧化铁皮中FeO层厚百分比减小,改善了机械剥离性能。

关键词: 72A高碳钢盘条, 剥离性能, 人工神经网络, 冷却制度, FeO层厚百分比