基于BP神经网络的螺杆泵井故障诊断方法

将神经网络原理应用于地面驱动螺杆泵采油系统的故障诊断,采用改进的BP神经网络,根据螺杆泵井的故障特点,通过理论研究,选取能够表征油井生产状态的状态变量作为神经网络的输入向量,归纳出常见的螺杆泵井故障形式作为目标输出。同时采集了大量现场数据,并进行分类整理,构成了网络的训练样本。通过对网络进行训练,获得具有一定泛化能力的网络。利用VB与Matlab编制的相应软件进行螺杆泵井的故障诊断,获得了正确的诊断结果,证明该方法具有一定的实用性。

Study of screw pump well fault diagnosis based on BP neural network

The principle of BP neural network (NN) is applied in the fault diagnosis of the vehicle drive screw pump exploitation system. Modified BP neural network is adopted. The state variables that can represent the production status of oil well are chosen by theory research on the failure characteristics of the screw pump well, taken as input vectors of the NN. Some common forms of failure are concluded, taken as the output vectors of the NN. Meantime, large amount of field data are collected, classified to form the training samples of the NN. Then the NN is trained by these collected data, a network that has good generalization ability is made. Two languages, VB and Matlab, are used to write the related program which is applied to the diagnosis, and correct results have been made. So it is proved that this kind of method has good practicability.