融合活塞杆轴心位置特征的往复压缩机多源信号故障诊断模型  

Multi-Source Signal Fault Diagnosis Model of Reciprocating Compressor Fused on the Axial Position Characteristics of Piston Rod

作　　者：冯泽江[1] 骆敏珠[1] 邓翔[1] 梁永龙 徐建平[1] FENG Ze-jiang;LUO Min-zhu;DENG Xiang;LIANG Yong-long;XU Jian-ping(Chuanzhong Oil and Gas Mine,PetroChina Southwest Oil and Gas Field Company,Suining 629000,China)

机构地区：[1]中国石油西南油气田分公司川中油气矿,四川遂宁629000

出　　处：《压缩机技术》2022年第4期21-28,33,共9页Compressor Technology

基　　金：重庆市技术创新与应用发展专项面上项目(cstc2020jscx-msxm0411);航空科学基金项目(ASFC-201834S9002)。

摘　　要：针对采用单一源数据的故障诊断方法难以适应往复压缩机故障诊断需求的问题,提出了一种融合活塞杆轴心位置特征的多源信号故障诊断模型。首先,提出一种改进的活塞杆轴心位置离散点外围轮廓提取方法,将二维空间坐标系中的离散点转化到极坐标系,通过坐标转换采用三次Hermite多项式插值法提取包络特征,利用仿真数据验证了轴心位置离散点外围轮廓提取新方法的优越性;然后,采用该方法快速提取活塞杆轴心位置离散点轮廓特征,包括离散点分布面积、包络熵等特征参数,以提升对活塞、活塞杆故障的感知能力;在基于自编码器网络自适应提取振动、位移信号特征基础上,融合活塞杆轴心位置离散点轮廓特征、振动信号机理特征、温度信号机理特征等,建立了一种多源数据特征融合的故障诊断模型;最后,利用故障模拟实验台实验数据与历史故障案例数据,对上述方法进行验证,结果表明:新的外围轮廓提取方法能够保证轴心轨迹外围轮廓提取效果并提升了处理效率;多源数据特征融合的故障诊断模型对阀片断裂、活塞杆断裂、活塞杆磨损、撞缸故障具有良好的诊断效果,同时验证了不同故障与多源信号数据之间的关联性。Aiming at the problem that the fault diagnosis method based on single-source data could not meet the requirements of reciprocating compressor fault diagnosis, a multi-source signal fault diagnosis model incorporating piston rod axis position characteristics was proposed.Firstly, an improved peripheral contour extraction method for discrete points of piston rod axis position was proposed, which converted the discrete points in a two-dimensional space coordinate system into a polar coordinate system, extracted the envelope characteristics utilizing three Hermite polynomial interpolation method through coordinate transformation, and verified the superiority of the new method for peripheral contour extraction of discrete points of axis position with simulation data.Secondly, the method was used to quickly extract the contour characteristics of discrete points of piston rod axis position, including the distribution area of discrete points, envelope Entropy, and other characteristics parameters, to improve the ability to sense piston and piston rod faults.Based on self-encoder network adaptive extraction of vibration and displacement signal characteristics, a fault diagnosis model based on multi-source data feature fusion was established by combining the characteristics of piston rod axis position discrete point contour, vibration signal mechanism, and temperature signal mechanism.Further, the above methods were validated by using the experimental data of the fault simulation laboratory and historical fault case data.The results show that the new peripheral contour extraction method could ensure the extraction effect of peripheral contour of axis track and improve the processing efficiency.The fault diagnosis model based on feature fusion of multi-source data has a good diagnostic effect on valve disc fracture, piston rod fracture, piston rod wear, and cylinder impact faults, and verifies the correlation between different faults and multi-source signal data.

关 键 词：多源数据融合 往复压缩机 活塞杆轴心位置 故障诊断 

分 类 号：TH457[机械工程—机械制造及自动化]