一种乳化液泵分级故障诊断方法

针对乳化液泵故障机理复杂、故障诊断难的现状,提出一种乳化液泵分级故障诊断方法。首先,通过深度自编码网络(Deep Auto-Encoder Network,DAEN)实现乳化液泵故障的第一级诊断,以乳化液泵的14个特征参数作为输入,自适应特征学习,识别故障形式;然后,通过专家系统实现乳化液泵故障的第二级诊断,将已识别的故障形式与必要故障信息作为专家系统输入,得到明确的故障定位。实验表明,深度自编码网络平均准确率98.712%,优于深度神经网络和卷积神经网络,可靠性高,可以完成第一级诊断任务,然后通过专家系统完成第二级诊断任务,分析产生原因,操作简单。将该方法编制成后台可运行的程序,嵌入煤矿综采工作面智慧云平台。经过实际测试,该故障分级诊断方法能够快速有效定位故障位置,提高诊断精度。

A Hierarchical Fault Diagnosis Method for Emulsion Pump

Aiming at the current situation of complex fault mechanism and difficult fault diagnosis of emulsion pump, a fault identification and location method for emulsion pump is proposed. Firstly, through a deep auto-encoder network (DAEN) realize the first level fault diagnosis of emulsion pump. Taking 14 characteristic parameters of emulsion pump as input, the fault forms are identified by adaptive feature learning. Then, through expert system realize the second level fault diagnosis. The identified fault forms and necessary fault information are used as input to receive a clear fault location. Experiments show that the classification accuracy for the deep auto-encoder network is 98.712 %, which is better than the comparable deep neural network and convolutional neural network. It has high reliability and can competent the first level diagnosis task. Throuh the expert system competent second level diagnosis task to analyze the cause, easy to operation. The method is programmed into a background program and embeded into the intelligent cloud platform of fully mechanized coal mining face. Through actual tests, the fault classification diagnosis method can quickly and effectively locate fault location and improve diagnosis accuracy.