加速度计信号奇异值分解滤波与参数辨识

Singular Value Decomposition Filtering and Parameter Recognition of Accelerometer Signals

针对石英挠性加速度计模型参数辨识的噪声影响问题,采用奇异值分解技术实现测量信号的滤波降噪处理,再结合总体最小二乘算法实现模型参数的高质量辨识。以重力场下的多齿分度台为测试分析平台,利用样本熵和统计参数定量评价辨识效果。试验结果表明,奇异值分解技术能有效去除信号中的噪声干扰,且通过降噪信号获得的模型系数稳定性较好。所采用方法有效地实现了加速度计参数辨识分析,可用于其性能质量评价。

Aiming at the noise influence problem in model parameter recognition of quartz flexible accelerometer,by adopting the singular value decomposition technology,the filtering noise reduction for measured signal is implemented; then high quality recognition of model parameter is realized by combining the total least square algorithm. With the multi-tooth indexing bench under gravitational field as the test and analysis platform,the recognition effects are evaluated quantitatively by using sample entropy and statistical parameters. The experimental result shows that the noise interference in signals can be effectively eliminated by singular value decomposition technology,and the model coefficients obtained from denoised signals are stable. The method proposed effectively implements parameter recognition analysis for accelerometer,it can be used in performance quality evaluation.