基于人工鱼群神经网络的抽油机故障诊断

摘要

BP神经网络应用于抽油机的故障诊断时易陷入局部极值,同时收敛速度也无法保证.在此前提下,提出人工鱼群神经网络算法的抽油机故障诊断新方法,充分利用人工鱼群在全局范围的快速寻优特性以克服BP神经网络收敛速度较慢和易陷入局部最优解的缺点,从而提高故障诊断的准确率和速度.以抽油机的管漏、供液不足、杆断脱、泵漏失、气影响五种故障类型为例,利用MATLAB分别搭建了传统BP神经网络和人工鱼群神经网络的模型,并对两种方法的诊断结果进行了比较.仿真结果充分说明了人工鱼群神经网络在抽油机故障诊断中的可行性、准确性和优越性.%BP neural network, when applied to fault diagnosis of oil pumping machines, will easily fall into local extremum, and convergence speed cannot be guaranteed.Under this premise, a new fault diagnosis method for oil pumping machines is presented on the basis of artificial fish swarm neural network algorithm to make full use of the algorithm's fast optimizing character on the global scope to overcome the shortcomings of BP neural network: slow convergence rate and likeliness to fall into locally optimal solution, thus improving the accuracy and speed of fault diagnosis.Taking five fault types (tube leakage of pumping machine, insufficient feed liquid, stem break-off, pump leakage and gas impact) as example, this paper uses Matlab to establish a model for traditional BP neural network and a model for artificial fish swarm neural network, and compares diagnosis results of these two methods.Simulation results sufficiently prove the feasibility, accuracy and superiority of artificial fish swarm neural network used for fault diagnosis of oil pumping machines.