基于神经网络的两层分采同步抽油系统井下故障诊断技术

Neural-Network-Based Down-Hole Fault Diagnosis Technique for Sucker-Rod Pumping System of Double-Layer Separate Recovery

建立了两层分采同步抽油系统的诊断模型 ,提出了求解这一非完整系统的迭代差分法 ,并由此建立了基于神经网络的两层分采同步抽油系统井下故障诊断专家系统。对专家系统各个模块的功能及其整个系统的运行机制进行了论述 ,阐明了 BP网络诊断模块的初始化、训练和测试过程 ,并采用带有自适应学习率和动量因子的改进 BP算法对其进行训练。借助功能强大的 MATLAB语言系统及其工具箱 ,完成了故障诊断专家系统软件的设计。

The existing fault diagnosis technique of sucker-rod pumping system can not meet the needs to diagnose the faults of sucker-rod pumping system of double-layer separate recovery. In this paper, the diagnostic model of sucker-rod pumping system of double-layer separate recovery is established, an iterative difference method of the noncomplete system is proposed, and then a neural-network-based down-hole fault diagnosis expert system of sucker-rod pumping system of double-layer separate recovery is built. The function of various blocks and the working principle of expert system is discussed, the initialization, training and testing process of BP network is illustrated, and the BP network is trained with the improved BP algorithm that has self-adapting learning rate and momentum factor. The software of the fault diagnosis expert system has been designed with the aid of the MATLAB and its Toolbox.