基于随机并行梯度下降算法的脉冲光纤放大器时域特性调控

Time-Domain Characteristic Regulation of Pulse Fiber Amplifier Based on Stochastic Parallel Gradient Descent Algorithm

提出了一种调控脉冲光纤放大器时域特性的新方法。根据光纤放大器中脉冲时域波形畸变的基本原理以及实际应用中对输出脉冲的具体要求,采用随机并行梯度下降(SPGD)算法获取生成目标输出脉冲所需要的最优输入脉冲。通过数值仿真,实现了多种指定脉冲的输出,证明了所提方法的有效性。利用SPGD算法可不断改变输入脉冲,使光纤放大器的真实输出逼近目标输出,该控制策略简单且输出脉冲收敛迅速。利用该方法获取直线边界脉冲(如矩形、M型和三角型脉冲)和曲线边界脉冲(如抛物线型和椭圆型脉冲)时均取得较好的效果,因此该方法适用于多种类型目标脉冲的生成。

A new method for regulating the tirne-domain characteristics of pulse fiber amplifier is proposed. According to the basic principle of pulse time-domain waveform distortion in fiber amplifier and the specific requirement in practical application of output pulse, the optimum input pulse required to generate target output pulse is obtained by the stochastic parallel gradient descent （SPGD） algorithm. Several particular output pulses can be generated in numerical simulation, which proves the effectiveness of the proposed method. The input pulse can be changed continually when we use the SPGD algorithm, so that the practical output of fiber amplifier comes close to the target output. The control strategy is easy and the output pulse converges rapidly. Good results can be achieved with the proposed method when we generate the pulses with linear boundaries such as rectangular, M shaped and triangle pulses, or generate the pulses with curved boundaries such as parabola and ellipse pulses, and the proposed method can be widely applied to generating variety types of target pulses.