针对水泵机组故障诊断复杂,提取的征兆和故障之间关系较为模糊,并且传统的BP算法存在易于陷入局部极小点、收敛速度慢、兼容性不好等问题,采用了一种动量因子与自适应学习率相结合的改进BP算法来建立神经网络模型,并通过有限的样本进行训练,最后通过几种故障特征向量进行验证。实际验证表明,此网络能够满足水泵机组故障诊断的技术要求。%The fault diagnosis of pump unit is complex ,and the relationship between the extracted symptoms and faults is fuzzy ,and the traditional BP algorithm is easy to fall into local minimum .In this paper ,an improved BP algorithm based on the combination of momentum factor and adaptive learning rate is used to establish the neural network model ,and it is trained through limited samples , finally ,several fault feature vectors are verified .The practical verification shows that this network can meet the technical requirements of the fault diagnosis of the pump unit .
展开▼