论文部分内容阅读
温度是影响光纤陀螺测试精度的重要因素,通过理论分析和实验研究了光纤陀螺的静态温度特性,提出采用经典小波网络进行零偏温度建模的方案,并与多项式拟合结果进行比较,其拟合精度得到大幅度提高。在此基础上,对经典小波网络进行改进,研究了一种新的参数初始化方法,并提出了基于动量变步长梯度下降法的参数更新算法。实验结果表明,改进的小波网络算法能够进一步提高网络的收敛速度和模型拟合的精度,从而能够更好地描述光纤陀螺零偏的温度特性。为了验证小波网络算法的普适性,利用多个惯性器件的温度实验数据,进行零偏温度建模和精度分析,结果表明,该方法能够较好地适用于对温度敏感的惯性器件的静态温度建模。
Temperature is an important factor that affects the test accuracy of fiber optic gyroscopes. The static temperature characteristics of the fiber optic gyroscope are studied theoretically and experimentally. A scheme of modeling the temperature of partial zero temperature using classical wavelet networks is proposed. The results are compared with the polynomial fitting results. The precision has been greatly improved. On this basis, the classical wavelet network is improved, a new parameter initialization method is studied, and a parameter updating algorithm based on the variable step size gradient descent method is proposed. The experimental results show that the improved wavelet network algorithm can further improve the network convergence speed and model fitting accuracy, and thus can better describe the temperature characteristics of FBG bias. In order to verify the universality of the wavelet network algorithm, the temperature experimental data of multiple inertial devices were used to model and analyze the zero-bias temperature. The results show that this method can be well applied to the static temperature of the temperature-sensitive inertial device Modeling.