基于杂合支持向量回归机的电子装备故障预测

Fault Prediction for Electronic Equipment Based on Hybrid Support Vector Regression

针对电子装备性能特征参数间的耦合关联问题,提出一种基于杂合支持向量回归机的电子装备故障预测方法。运用D-S证据理论,结合参数的纵向历史状态数据和横向的相关参数数据,设计杂合支持向量回归机预测算法,利用特征参数的时间相关性和空间相关性提高预测精度。实验结果表明,相对于单独使用纵向或者横向的支持向量回归机,该方法具有更高的精度,可有效地对复杂电子装备实施故障预测。

Aiming at the coupling relationship among electronic equipment＇s feature parameters, a hybrid Support Vector Regression（SVR） based fault prediction model is proposed in the paper, both the time relativity and space relativity of the feature parameter are taken into account, and it designes the algorithm process of hybrid SVR which improves the prediction accuracy by fusing the vertical historical state data and the horizontal correlation parameter together, applying D-S evidence theory. Experimental results show that, compared with vertical SVR or horizontal SVR, the proposed method is more accurate, and is capable of performing fault prognosis on the complicated electronic equipment effectively.