基于相关向量机的氢气传感器故障恢复方法

Fault Recovery Method of Hydrogen Sensor Based on Relevance Vector Machine

在相关向量机回归模型的基础上,提出了一种新的氢气传感器故障数据恢复方法。利用小生境粒子群算法的"共享机制"对相关向量回归的核参数进行了优化,使其能快速准确地找到全局最优参数。用优化并训练后的回归模型对发生故障后的数据进行预测,实现故障恢复。将本文所用方法与其他较成熟的方法进行了比较,实验结果表明本方法在恢复准确度和鲁棒性方面均优于传统方法。数据恢复相对误差在±2.8%以内。

Based on relevance vector machine regression model, a new fault data recovery method of hydrogen sensor was pro- posed in this paper. The ＂sharing mechanism＂ of niche particle swarm optimization algorithm was used to optimize kernel parame- ter of PVM ,which can make it to find the global optimal parameter fast and exactly. The fault data were prognosed by using the regression model of optimized and trained to realize the fault recovery. The proposed method was compared with other mature algo- rithms, the results show that the proposed method is superior to the traditional ones in recovery accuracy and robustness.The relative error of fault recovery is within ＋2.8%.