基于改进的最小二乘支持向量机的退火炉内带钢延伸量预测

带钢在退火炉内进行长时间加热的过程中，由于炉内温度、张力以及带钢运行速度等变量间具有多重相关性且存在生产数据噪声，从而难以准确预测炉内带钢的延伸量。本文提出了一种基于遗传算法的BP神经网络理论(GA-BP神经网络)与最小二乘支持向量机(LSSVM)理论相结合的算法用于目标函数拟合，该算法兼顾最小二乘支持向量机的全局与局部拟合能力，有效避免算法陷入局部最优的缺点，利用改进的BP神经网络算法优化选择LSSVM 模型的惩罚因子和核函数参数，以避免人为选择参数的盲目性，使LSSVM具有更好的泛化能力和预测精度。将该方法应用到退火炉内带钢的延伸量预测，通过现场生产数据仿真模拟进行非线性函数估计，结果表明：本文所提出的方法具有良好的预测精度，可以应用于实际生产。

Application of soft sensor in elongation prediction of strip steel based on improved LSSVM algorithm

The strip in annealing furnace endures long time heating,so the strip elongation is hard to predict as for data noise and multi-correlation between furnace temperatures,tension and strip speed.A BP neural network algorithm based on genetic algorithm (BP-GA) combining with least square support vector machine (LSSVM) is used for objective function fitting.The proposed algorithm has the global fitting and local fitting characteristics to avoid local optimum.This improved BP neural network is applied for selection of penalty factor and kernel function to avoid blindness of artificial selection,and make the LSSVM get better generalization and prediction accuracy.An actual simulation for the elongation prediction of strip steel in the annealing furnace,with the nonlinear function estimation,shows that this proposed algorithm has good prediction accuracy and practical value.