论文部分内容阅读
随着光纤承载业务量的逐渐增加,光纤线路产生的故障数据越来越多,传统方法对光纤故障数据的采集过程复杂,不能快速得到故障信息,导致光纤故障监测实时性能差,无法及时发现故障,因此,提出基于改进神经网络的光纤故障监测方法,通过小波变换法对光纤OTDR曲线进行分析,得到光纤信号中的所有细节,获取故障光纤信号,利用小波包对故障光纤信号进行特征提取,将提取到的光纤故障向量作为神经网络的输入,通过前向计算过程、误差计算和误差反向传播过程完成神经网络的训练。针对BP神经网络收敛速度慢和学习率、惯性系数确定方法不合理的弊端,利用自适应学习速率动量梯度下降反向传播算法对神经网络进行改进,给出利用改进方法对光纤故障进行监测的实现过程。实验结果表明,所提方法具有很高的故障监测精度。
With the gradual increase in the carrying capacity of optical fibers, more and more faulty data are generated on the optical fiber lines. The traditional method for collecting optical fiber fault data is complex and can not obtain fault information quickly, resulting in poor real-time performance of optical fiber fault monitoring and failure to discover faults in time Therefore, this paper proposes an optical fiber fault monitoring method based on improved neural network. The optical fiber OTDR curve is analyzed by wavelet transform to get all the details of the optical fiber signal. The faulty optical fiber signal is obtained. The wavelet packet is used to extract the feature of the faulty optical fiber signal. The extracted fiber fault vector is used as input of neural network, and neural network training is completed through forward calculation process, error calculation and error back propagation process. Aiming at the disadvantages of slow convergence speed, unreasonable learning rate and inertia coefficient of BP neural network, this paper improves the neural network by using the adaptive learning rate gradient descent backpropagation algorithm, and presents an improved method to monitor the optical fiber fault process. Experimental results show that the proposed method has high fault monitoring accuracy.