L-M神经网络算法在汽轮发电机组故障诊断中的应用

Application of Artificial Neural Network Based on Levenberg-Marquardt Algorithm to Fault Diagnosis of Turbine Generator Set

针对汽轮发电机组的故障诊断,采用Levenberg-Marquardt算法建立多层前向人工神经网络,采用改进算法训练网络,克服了传统BP算法收敛速度慢,易陷入局部最小的缺陷.就BP网络的不足,提出了一种改进的BP神经网络模型,并使用L-M算法用于汽轮发电机组故障的诊断.经理论和实践证明:该方法有效地提高了故障诊断的精度和可靠度,为旋转机械故障诊断提供了有效方法.

An artificial neural network（ANN） using Levenberg-Marquardt algorithm for network training is presented to diagnose faults in turbine generator set equipment. The ANN uses improved training method to avoid the drawbacks of BP algorithm, and diagnoses faults by L-M algorithm. Proved theoretically and practically,this method increases efficiently the precision and reliability of breakdown diagnosis, and provides efficient way for breakdown diagnosis of whirling machinery.