钻井泵液力端故障诊断新方法-石油学报.pdfVIP

3 0 4 Vol. 3 0 No. 4 2 009 7 A CT A PETROLEI SINICA July 2 00 9 : 0253-2697 (2 00 9) 04-0 617-04 1 2 3 4 裴峻峰 张嗣伟 齐明侠 万广伟 ( 11 2 13016; 21 102249; 31 257061; 41 473 132) : 往复泵液力端故障原因及故障与征兆间对应关系复杂, 为了全面地利用获取的振 信号资源, 得到更全面准确的诊断结果, 将分析得到的幅值域的峭度指标峰值指标脉冲指标裕度指标波形指标和歪度等6 个参数, 频域的重心频率均方根频率频率 标准差等3 个参数以及32 个小波包分频带能量值作为神经网络输入的备选特征向量, 由此形成了液力端综合振 信号特征参数 的神经网络诊断系统为了对网络的性能进行比较, 分别构建了BP 网络和RBF 网络将上述特征输入向量作不同组合, 分别输 入该网络并进行训练诊断和效果对比, 由此求得了最优诊断系统组合利用此神经网络诊断系统, 对现场实际使用的钻井泵液力 端进行了多次的测试分析和调试, 证明这种方法对钻井泵液力端的故障诊断是行之有效的, 可取得较高的诊断准确率 : 钻井泵; 液力端; 故障诊断; 振 信号; 特征参数; 神经网络诊断系统 : T E825 : A A new method for fault diagnosis of fluid end in drilling pump 1 2 3 4 PEI Junfeng ZHAN G Siw ei QI Mingxia WAN Guangw ei (1. College of Mechanical and Energy Engineering, Jiangsu Polytechnic University, Changzhou 2 13 016, China; 2. China University of Petroleum, Beij ing 1022 49, China; 3. College of Mechanical and Electronic Engineering, China University of Petroleum, ongying 2570 61, China; 4. rilling Eng ineering Comp any , H enan Petroleum Exp loration Bureau, N anyang 4 73 132, China) Abstract: The faults are resulte from some complex reasons in flui en of rilling pump. The correspon ing relationship bet ween faults an symptoms is complex . In or er to totally use information resources of vibration signal an to obt ain more comprehensive an precise results, six amplitu e- omain in exes such as Kurtos