摘要
文章基于CNN和LSTM构建多重轨道交通电力电缆故障类型诊断模型。首先介绍了神经网络的基本理论,分析了其局限性,并针对这些局限性总结了一些神经网络常用的优化方法;采用参数优化算法对多重神经网络的初始权值和阈值进行优化,并通过四个标准测试函数对粒子群算法进行了收敛性测试。最后建立CNN-LSTM模型实现对轨道交通电力电缆故障诊断。经验证,多重神经网络在算法稳定性、正确率和快速性上都有优势。
The article constructs multiple types of rail transit power cable fault diagnosis based on CNN and LSTM.Firstly,the basic theory of neural networks was introduced,and their limitations were analyzed.In response to these limitations,some commonly used optimization methods for neural networks were summarized.The initial weights and thresholds of multiple neural networks were optimized using parameter optimization algorithms,and the convergence of the particle swarm optimization algorithm was tested using four standard test functions.Finally,a CNN-LSTM model was established to diagnose power cable faults in rail transit.After verification,multiple neural networks have advantages in algorithm stability,accuracy,and speed.
作者
张辉
ZHANG Hui(Wuxi University,Wuxi 214000,China)
出处
《数字通信世界》
2023年第8期127-129,共3页
Digital Communication World
关键词
轨道交通电力电缆
模态分解
混合神经网络
rail transit power cable
modal decomposition
hybrid neural network
作者简介
张辉(2002-),男,汉族,江苏兴化人,本科在读,研究方向为人工智能。
