基于MLP神经网络的海上固井泵故障诊断方法  

Fault diagnosis method of offshore cementing pump based on MLP neural network

作　　者：张冉[1] 赵琥[1] 侯林[1] 陈峰 邵振友 季威 蔡宝 ZHANG Ran;ZHAO Hu;HOU Lin;CHEN Feng;SHAO Zhenyou;JI Wei;CAI Bao(China Oilfield Services Limited,Langfang,Hebei,065201 China)

机构地区：[1]中海油田服务股份有限公司,河北廊坊065201

出　　处：《自动化与仪器仪表》2022年第9期54-57,共4页Automation & Instrumentation

基　　金：中海油服油化EMP系统推广实施(G2115A-0521C122)。

摘　　要：海上固井泵受到高温高盐环境因素影响,导致故障发生率较高,为了提高海上固井泵的故障检测诊断能力,提出基于MLP神经网络的海上固井泵故障诊断方法。构建海上固井泵的故障传感信息采集模型,采用相关性参数分析方法,进行海上固井泵故障数据特征融合和自适应参数解析,构建海上固井泵故障远程协作诊断的特征分析和数据分析模型,结合海上固井泵故障参数的多元耦合分析结果,通过关联规则挖掘的方法分析海上固井泵的故障特征量,采用MLP神经网络学习方法,实现海上固井泵振动传感信息融合及滤波成分分析,结合关联特征挖掘和模糊信息聚类,建立海上固井泵的故障类别参数融合和信息聚类模型,通过对海上固井泵的振动传感异常特征分析,采用MLP神经网络实现对海上固井泵的故障类型化参数跟踪识别,根据信息聚类结果,实现故障分类检测和诊断。仿真结果表明,采用该方法进行海上固井泵故障诊断的准确率较高,达到了0.97,时间开销平均为3.73 s,提高了海上固井泵的工况稳定性。Offshore cementing pump is affected by high temperature and high salinity environment factors, which leads to high failure rate. In order to improve the fault detection and diagnosis ability of offshore cementing pump, a fault diagnosis method of offshore cementing pump based on MLP neural network is proposed. The fault sensing information collection model of offshore cementing pump is constructed, the correlation parameter analysis is used to carry out the feature fusion and adaptive parameter analysis of offshore cementing pump fault data, and the feature analysis and data analysis model of offshore cementing pump fault remote cooperative diagnosis is constructed. Combining with the multi-coupling analysis results of offshore cementing pump fault parameters, the fault feature quantity of offshore cementing pump is analyzed by association rule mining method, and MLP neural network learning method is adopted. Realize the vibration sensing information fusion and filter component analysis of offshore cementing pump, combine the correlation feature mining and fuzzy information clustering, and establish the fault category parameter fusion and information clustering model of offshore cementing pump. By analyzing the abnormal characteristics of vibration sensing of offshore cementing pump, realize the tracking and identification of the fault type parameters of offshore cementing pump, and realize the fault classification detection and diagnosis according to the information clustering results.The simulation results show that the accuracy of fault diagnosis of offshore cementing pump by this method is high, reaching 0.97, and the average time cost is 3.73 s, which improves the working stability of offshore cementing pump.

关 键 词：MLP神经网络 海上固井泵 故障 诊断 关联特征挖掘 模糊聚类 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]