为了准确识别激光焊接的稳态类型, 采用了图像处理、相关分析和神经网络的方法, 增加对准稳态的研究, 以信号特征的相关系数作为神经网络模型的输入, 进行了理论分析和实验验证, 得出了光学、视觉信号的相关性对激光焊接稳态类型的影响规律。结果表明, 匙孔面积和金属蒸汽面积的相关性区分稳态类型的效果最好, 其相关系数为0.2~0.3时为稳态, 0.4~0.5时为准稳态, 0.6~0.7时为非稳态; 训练完成的神经网络模型在测试集上达到了98.76%的预测准确率, 可满足准确识别焊缝稳态类型的需求。该研究为自动化生产中预防出现激光焊接缺陷提供了参考。

相干光正交频分复用系统（CO-OFDM）是未来高速光传输的重要解决方案，而偏振模色散（PMD）与偏振相关损耗（PDL）相互作用严重影响传输性能。分析了CO-OFDM系统中的偏振效应和数学模型，并通过仿真加以验证。结果表明，一阶PMD效应对正交频分复用(OFDM)信号子载波幅度附加余弦因子；二阶PMD效应中，去偏振项（DR）占主要因素，且一定程度上可缓解色度色散带来的相位噪声；另外，由于PMD带来OFDM子载波偏振度随频率变化加强，也对PDL带来的系统损伤有一定抑制作用，在不考虑色度色散，相对损耗因子α=0.5，单模光纤传输距离为720 km条件下，系统Q值约有2 dB提高。

A new method incorporating biased clipping orthogonal frequency division multiplexing (OFDM) is presented, which mitigates fiber nonlinear effects in a long-haul coherent optical OFDM (CO-OFDM) system. Under the scheme of the method, the wanted signal carried by odd subcarriers is orthogonal to clipping noise and a Mach-Zehnder modulator (MZM) performs the optimal OFDM signal up-converter from the radio frequency (RF) domain to the optical domain. Analysis and simulation results show that fiber non-linear effects can be effectively mitigated by reducing the peak-to-average power ratio (PAPR) in biased clipping CO-OFDM system. The nonlinearity threshold (NLT) is improved by 5 dB with a reach of 240 km. With a fiber length up to 800 km, system Q value is improved by approximately 2.3, 1.2, and 0.6 dB at a chromatic dispersion of 6, 12, and 16 ps/(nm￠km), respectively. Additionally, system Q reaches the maximum when direct currect (DC) bias is equal to the mean value of the OFDM waveform.

为了使系统的总功率达到最小,研究了OFDM子载波分配算法,采用遗传算法分配子载波。提出了一种有效的交叉算法并且将遗传算法进行了改进,有效保证每个用户的传输,改进后优于其它算法,能达到资源分配的最佳效果。它易于实现,效果明显,它比其它传统的优化算法有更优越的性能。