基于自适应RVM的电子系统缓变故障预测方法

针对电子系统缓变故障的预测问题,提出一种自适应相关向量机(RVM, Relevance Vector Machine)方法.首先,对反映电子系统性能的参数序列进行相空间重构,建立RVM的输入输出对应关系;然后,将嵌入维数和核函数参数作为人工鱼位置,取留一交叉验证(LOOCV, Leave-One-Out Cross-Validation)误差的相反数作为目标函数,利用人工鱼群算法(AFSA, Artificial Fish Swarm Algorithm)实现方法参数的自适应优化选择;最后,通过雷达发射机高压电源与多注速调管的故障预测实验验证了方法的性能.实验结果表明:该方法在预测精度和预测可靠性方面优于现有方法.

Gradual fault prediction method for electronic system based on adaptive RVM

Aiming at gradual fault prediction of electronic system, a method based on adaptive relevance vector machine (RVM) was proposed. Firstly, the phase space of electronic system performance parameter time series was reconstructed, and then the corresponding relationship between the input and the output of RVM was educed. Secondly, the artificial fish swarm algorithm (AFSA) was adopted to realize the adaptive optimization of method parameters, which took the embedding dimension and the RVM kernel parameter as the artificial fish position and took the opposite number of leave-one-out cross-validation (LOOCV) error as the objective function. Lastly, the performance of the proposed method was validated by radar transmitter fault prediction experiment. The experimental results show that the proposed method has better accuracy and reliability than the existed methods.