基于在线BP神经网络的制冷系统风机故障识别

Fan fault detection in refrigeration system based on online BP neural network

本文介绍了一种基于神经网络在线学习的制冷系统风机故障识别方法。选取环境温度、蒸发温度、风机风量、压缩机负荷作为输入,以预测冷凝温度为输出,利用BP(Back Propagation,反向传播)神经网络对一台风冷冷水机组进行在线状态学习,并利用学习到的模型对冷水机组进行风机故障识别。将整个过程分为初始化,成长学习、成熟学习、暂停4个状态,介绍了每个状态的主要任务,及各状态如何过渡衔接。结果表明,该方法可以成功地应用在风冷冷水机组的在线实时控制中,对运行中风机的故障进行识别和报警。

Introduces a fan fault detection method for refrigeration system base on online machine learning.By taking ambient temperature,evaporation temperature,fan air flow and compressor load as input,and taking condenser temperature as output,the status model of an air cooled chiller was learned by the online type BP neural network.Fan fault is identified by using the model obtained from BP network.Four states of learning is defined:initialization,growth learning,maturity learning and pause.The main tasks of each state,and the transition and connection between the states is introduced.The results show that the method can be successfully applied in the on-line control of the air cooled chiller,and the fan fault can be successfully detected and alarmed.