线性神经网络在加速度计静态模型标定的应用

Application of linear neural network in calibration of accelerometer

研究了线性神经网络在加速度计静态误差模型参数标定中的应用。根据加速度计静态数学模型的特点,选用网络结构简单、收敛速度快而不存在局部极小值问题的线性神经网络对加速度计静态模型进行辨识。同时考虑了加速度计安装到分度头上标定时,加速度计的安装误差角和分度头的转动精度对标定精度影响,并给出了相应的误差角计算和补偿方法,有效地提高了加速度计静态误差模型的标定精度。试验表明线性神经网络是一种有效的加速度计静态模型的标定方法而且具有较高的精度。

Application of linear neural network in calibration of static error model of accelerometer is studled. According to the characteristics of accelerometer static model, the linear neural network with simple network structure and rapid convergence speed, which will not come into local minimum, is chosen for parameter calibration of static mathematical model of accelerometer. The errors resulted from installation of accelerometer are also taken into consideration, and the corresponding error calculation and compensation methods are presented. Thus the calibration accuracy of the accelerator is improved effectively. Practical experiment showed that linear neural network is an effective method for calibrating the static mathematical model parameters.