【摘要】 为精确诊断机械采油井的系统故障,提出一种基于小波包特征向量的神经网络故障诊断方法。对采集到的悬点位移和悬点载荷数据进行3层小波包分解,构造小波包特征向量,并以此为故障样本对3层BP神经网络进行训练,考虑到传统BP算法的局限性,采用改进的遗传算法训练网络权值,进而实现油井系统的智能化故障诊断。试验结果表明,训练好的神经网络能够很好地诊断出采油井故障类型。更多还原

【Abstract】 To diagnose accurately the system failure of mechanical production wells,a failure diagnosis method of neural networks based on wavelet packets energy eigenvector was proposed.It adopted three-layer wavelet packets to decompose the collected data of suspended displacement and load,and construct the wavelet packets energy eigenvectors,then take these wavelet packets energy eigenvectors as failure samples to train three-layer BP(Back Propagation) neural network.In considering the limitation of traditional BP algorithm,an improved genetic algorithm was adopted in this paper to train the network weights,which implemented intelligent fault diagnosis of well systems.The experimental results show that the trained neural network can diagnose the failure types of production wells.更多还原