基于改进支持向量回归的港机剩余使用寿命预测

为解决标准支持向量回归(support vector regression, SVR)在预测港机剩余使用寿命时,误差程度不同的数据会被施以相同程度的惩罚,从而导致预测精度不足的问题,通过对每个数据的惩罚系数添加各自的误差比例,使每个数据的惩罚系数与各自的误差程度成比例,降低误差数据对整体预测精度的影响。实验表明,这种改进SVR相对于标准SVR在不降低计算效率的前提下能有效提高预测精度,并拥有良好的泛化能力。

Prediction of remaining useful life of port machine based on improved support vector regression

When the standard support vector regression (SVR) is adopted to protect the remaining useful life of port machine, the data with different degrees of error are given the same degree of punishment, which leads to insufficient prediction accuracy. In order to solve the problem, the respective error ratio is added to the penalty coefficient of each data, so that the penalty coefficient of each data is proportional to the respective error degree, and the influence of the error data on the overall prediction accuracy is reduced. Experiments show that, this improved SVR can effectively improve the prediction accuracy and have good generalization ability compared with the standard SVR without reducing the computational efficiency.