提出一种基于人工神经网络(ANN)和粒子群优化 (PSO) 的半导体激光器参数反向设计方法。利用由传统数值仿真方法计算出的激光器功率样本数据来训练ANN,并用此网络预测激光器任意一组参数对应的功率谱,均方差可低至0.5 mW,用时仅0.07 s,计算速度提高了约1800倍(与相同环境下传统数值算法耗时125.57 s相比)。将此网络与PSO算法结合,可获得目标功率谱的对应参数,即实现反向设计。经计算获得的反向设计方案不唯一,从而进一步验证了半导体激光器非线性多参数的特点。相同环境下ANN结合PSO的反向算法(均方差低于0.04 mW,用时39.45 s)与传统数值反向方法(均方差为0.89 mW,用时192 h)相比,精度提高了22.25倍,速度提高了约17500倍,说明了该方法的有效性。

This study proposes a novel semiconductor laser parameter inverse design method based on artificial neural network (ANN) and particle swarm optimization (PSO) algorithm. The ANN is trained using a laser output power as sampling data, which can be calculated by applying the traditional numerical simulation method. The network can be used to predict the power spectrum of the laser for any new values of the selected parameters. The mean square error can be as low as 0.5 mW and the CPU time as low as 0.07 s, which is about 1800 times more efficient than that of the numerical algorithm, which takes 125.57 s CPU time in the same environment. To obtain the design parameters for any target power spectrum, the inverse design can be achieved by combining this network with the PSO algorithm. It is clear from the calculation that the inverse design parameters are not unique, which proves that the semiconductor laser has a nonlinear multi-parameter problem. The combination of ANN and PSO inverse algorithm (with a mean square error of less than 0.4 mW and a CPU time of 39.45 s) demonstrates greater performance based on the same condition when compared with the traditional numerical simulation inverse method (with a mean square error of less than 0.89 mW and CPU time of 192 h). The accuracy and speed of the proposed method are improved by 22.25 times and about 17500 times, respectively.