论文部分内容阅读
以较成熟的ＢＰ（ｂａｃｋ－ｐｒｏｐａｇａｔｉｏｎ）神经网络为基础，论述了神经网络技术在旋转机械故障诊断中的应用问题，提出了用并行ＢＰ网络解决并发故障分类问题的方法，并采用自适应学习率和绝对误差等距离逼近法改善ＢＰ网络的学习性能，最后以振动频谱为特征，就汽轮发电机组机械中常见的２０种故障模式的分类和训练学习进行了仿真试验和分析。结果表明：并行ＢＰ网络可以解决并发故障分类问题，为旋转机械智能诊断提供了一个有效方法
Based on the more mature BP (back-propagation) neural network, the application of neural network technology in the fault diagnosis of rotating machinery is discussed. A parallel BP network is proposed to solve the problem of concurrent fault classification. Adaptive learning rate And absolute error equal distance approximation method to improve the learning performance of BP network. At last, the vibration frequency spectrum is taken as the characteristic to carry on the simulation test and analysis on the classification and training learning of the common 20 fault modes in the turbine generator set. The results show that parallel BP network can solve the problem of concurrent fault classification and provide an effective method for intelligent diagnosis of rotating machinery