论文部分内容阅读
环境温度变化造成的复杂漂移(温度漂移)始终是制约光纤陀螺(FOG)性能提高的重要因素。FOG温度漂移本质上是一组与温度有关的多变量非线性时间序列。在这一领域首次采用投影寻踪学习网络(PPLN)方法设计FOG温度漂移在线估计器。相对于传统的神经网络技术,PPLN采用批量学习和参数交替优化的训练算法,可以自适应确定神经网络的规模、参数和神经元函数,不仅具有简捷的网络结构和较强的鲁棒性和模型辨识能力,还可以有效克服学习过程局部极限问题。基于该方法设计PPLN漂移估计器对某型FOG温度漂移进行估计。采用试验实测数据对所提方法进行验证,并采用传统反向传播神经网络(BPNN)的方法进行比较,计算分析结果表明,PPLN漂移估计器具有更好的估计精度和鲁棒性,尤其在陀螺温度不正常变化时对当前漂移的估计精度可以提高至少2倍。
Complex drift (temperature drift) caused by changes in ambient temperature is always an important factor that restricts the performance improvement of FOG. FOG temperature drift is essentially a set of temperature dependent multivariate nonlinear time series. For the first time in this area, the POG temperature drift online estimator is designed using the Projection Pursuit Learning Network (PPLN) approach. Compared with the traditional neural network technology, PPLN adopts the training algorithm of batch learning and parameter alternation optimization, and can adaptively determine the size, parameters and neuron functions of the neural network, not only has a simple network structure and strong robustness and model Recognition ability, but also can effectively overcome the learning process of the local limit problem. Based on this method, a PPLN drift estimator is designed to estimate the temperature drift of a FOG. The experimental data are used to verify the proposed method and compared with the traditional backpropagation neural network (BPNN) method. The results show that the PPLN drift estimator has better estimation accuracy and robustness, especially in the gyro The estimation accuracy of the current drift can be increased by at least 2 times when the temperature changes abnormally.