Abstract: A multi-input and multi-outpul support vector regression algorithm is introduced to predict the real-time object surface temperature of casting slab in secondary cooling zone by using the given object teinperaiure schedule calculated by metallurgical technicians. Test results of casting 200 mm x 1 534 mm slab of steel 16Mn show that with same trainingspecimen number, the training time by support vector regression algorithm is 3. 2 s with error of predict object temperature ± 1 ℃ , while the training time by BP ( back propagation) neural networks algorithm is 23. 5 s with error of predict object temperature ± 2 ℃ , and the multi-input and output support vector regression aigorithm is better than BP neural networks algorithm to alter the object surface temperature of slab in real time according to the variety of process, therefore it is available to dynamic-control the casting process and increase the quality of slab.

Key words: Object Surface Temperatures of Slab, Secondary Cooling, Support Vector Regression

摘要： 提出了一种多输入多输出支持向量机回归算法,利用冶金技术人员计算的目标温度设定表,设定实时二冷区铸坯表面目标温度。200 mm×1 534 mm 16Mn钢板坯连铸试验结果表明,在训练样本相同时,支持向量机训练时间为3.2 s,预测目标温度误差为±1℃,BP神经网络训练时间为23.5 s,预测目标温度误差为±2℃,多输入多输出支持向量机回归算法优于BP神经网络算法,能够根据工艺变化情况,实时改变目标温度,为实现连铸动态控制提供了条件,有助于提高铸坯的质量。

关键词: 铸坯表面目标温度, 二次冷却, 支持向量机回归