基于EEMD-RVM的陀螺漂移混合建模预测

陀螺漂移序列具有非平稳和非线性的特点,针对单一模型难以对其实现精确预测的问题,提出一种基于集合经验模态分解(EEMD)和相关向量机(RVM)的混合建模方法,实现对陀螺漂移序列的区间预测.首先,利用集合经验模态分解将漂移序列分解为多个模态和一个余量;将模态区分为噪声和趋势两个分量,对噪声分量建立分布模型,对趋势分量建立RVM模型,两者等权相加还原得混合模型;最后,给定置信度,得到置信区间预测结果.将该方法用于某振动陀螺漂移序列预测实例,结果表明:该混合预测模型能准确预测陀螺漂移,其中RVM的预测精度达到99.86%,且验证集以给定的置信度落在预测区间内,可为陀螺的寿命预测和性能分析提供依据.

Mixed modeling for gyro drift prediction based on EEMD-RVM

In view that the timeseries of gyro drift cannot be preciselypredicted by single forecasting model due to its non-linear and non-stationary characteristics,interval forecasting for gyro drift series can be obtainedwith hybrid modeling method based on ensemble empirical mode decomposition(EEMD)and relevance vector machine(RVM)which is proposed. Firstly,the drift data is decomposed into a series of intrinsic mode function and one residue via EEMD. Secondly,modes areclassified into two categories:noise component and trend component,the distributionmodel of noise component and the RVM model of trend componentis established,two models are added with equal weight to establish the hybrid model. In the end,we set the confidence coefficient to obtain interval forecasting. By using the proposed method for a vibratory gyro drift prediction,the experiment result shows:in hybrid model,RVM prediction accuracy is 99.86%,validation set is contained by prediction interval with designated confidence coefficient. The hybrid model could provide reliable evidence for life prediction and performanceanalysis of gyro.