In order to diagnose the elevator fault stop precisely, and according to the characteristics and structure of the intelligent elevator detecting system, this paper applies a method of three-layer BP neural network based on multi-sensor data fusion.For different sensor signals using different approach, training samples includes wavelet packet energy eigenvector, also involves such as kurtosis, peak to peak value and noise signal.Take the above eigenvectors as fault samples, the Quasi-Newton BP trained by the 104-step, and the numerical accuracy is 2.6× 10-4.The trained network can exactly diagnose the elevator fault stop.The results show that the Quasi-Newton BP algorithm is batter than the resilient back BP algorithm.%为了判断电梯运行是否故障急停,结合电梯动态智能检测系统的结构和特点,采用基于多传感器数据融合技术的3层BP神经网络方法,并将其应用到电梯动态智能检测系统中;因对不同传感器采集的信号采用不同的处理方法,训练样本包括基于小波包分析的能量特征向量,峭度系数、峰峰值时域特征值;Quasi-Newton BP算法经104步完成对样本训练,精度是2.6x10-4,实现检测系统的智能化急停诊断;结果表明该算法优于弹性BP算法.
展开▼