基于神经网络的分段供电驱动脉冲信号识别方法  

Method of segmented-powered system driving pulse signal analysis based on neural network

作　　者：徐兴华[1] 李文禄[1] 腾腾[1] 王洁[1] 

机构地区：[1]海军工程大学舰船综合电力技术国防科技重点实验室,武汉430033

出　　处：《海军工程大学学报》2017年第4期1-8,共8页Journal of Naval University of Engineering

基　　金：国家自然科学基金资助项目(51477178);国家重点基础研究发展计划资助项目(2013CB035601)

摘　　要：针对分段供电驱动信号的质量评价与边界识别问题,首先采用内积运算作为两组布尔信号相似性的度量给出了基于标准信号的测量矩阵,并提出了一种对样本权重平衡的方法训练神经网络,实现了对驱动脉冲频率和占空比异常的检测;然后,给出了抗干扰的驱动脉冲边界的识别方法;最后,通过实验对该方法进行了验证。实验结果表明:该方法可对驱动脉冲的信号质量和边界进行有效识别。Focused on the quality evaluation and boundary identification problem for segmentedpowering driving signals,firstly the inner product operation is used as the similarity metric of two Boolean signal set,and the measurement matrix based on the standard signal is given.Meanwhile,a sample weight balancing method is proposed for training the neural network to realize the detect of the abnormal of the frequency and the duty ratio.Then a recognition method of driving pulse boundary is proposed.Finally,the method is verified by the experiments.The results show that the method can effectively identify the quality and boundary of the drive pulse signal.

关 键 词：分段供电 驱动脉冲 占空比 测量矩阵 神经网络 

分 类 号：TM359[电气工程—电机]