摘要
针对传统功率谱信号源不足以及BP神经网络收敛速度慢且容易陷入局部极小等问题,提出矢功率谱和蚁群神经网络相结合的故障诊断方法,该方法是:提取矢功率谱的8个频段能量特征,并输入到蚁群神经网络分类器进行故障识别,通过实际训练结果和实验结果对比可知,蚁群神经网络能有效地提高收敛速度,网络迭代次数明显改善,故障识别率提高,将蚁群神经网络应用于机械故障诊断是有效的。
The signal source of traditional power spectrum is insufficient, the convergence speed of BP neural network is slow and may inevitably meet local minimal problems. According to these problems, a new fault diagnosis approach is proposed, this ap-proach is that the vector power spectrum is used as eigenvectors, the ant colony neural network as a classifier. The experiment re-suits shows that the ant colony neural network can effectively improve the convergence speed and fault identification. So the pro-posed approach applied to machinery fault diagnosis is very effective.
出处
《现代制造工程》
CSCD
北大核心
2013年第6期121-125,共5页
Modern Manufacturing Engineering
基金
国家自然科学基金青年科学基金项目(51205371)
河南省科技攻关计划项目(122102210122)
关键词
矢功率谱
蚁群算法
BP神经网络
故障诊断
vector power spectrum
Ant Colony Algorithm（ACA）
BP neural network
fauh diagnosis
