用多项式预测滤波消噪的传感器动态特性辨识

Identification of sensor's dynamic characteristic based on denoising by polynomial FIR predictors

在利用传感器进行动态测量时,为了得到精确的测量结果,需要建立传感器动态特性的数学模型,传感器动态特性可以通过系统辨识得到.但是,测量噪声的存在,使得辨识得到的传感器动态特性与实际动态特性存在一定误差,影响到测量系统的精度.为了解决该问题,本文讨论了多项式预测滤波和中值滤波相结合的方法对传感器输出信号进行滤波消噪.然后,利用消噪后的信号,通过系统辨识方法建立传感器动态特性的数学模型.研究表明,采用本文研究的方法可以克服测量噪声对传感器动态特性辨识的影响,并将该方法用于薄膜热电偶的动态特性辨识.

In dynamic measurements, it is necessary to build a mathematical model for a sensor＇s dynamic characteristic to obtain accurate measurement data, and the sensor＇s dynamic characteristic can be determined by system identification. However, the noises in measurement may produce adverse influences to system identification, and cause error between the identified characteristic and the real one. To reduce these disadvantageous influences, sensor output signal polluted by additive noise are processed by hybrid filter with polynomial FIR predictors and median （PMH）. After the signal is denoised by PMH, the mathematical model of sensor＇s dynamic characteristic is obtained by system identification. The study shows that the presented approach can reduce the influences of noise. This approach has been applied to the identification of the characteristic of a thin-film thermocouple.