基于控制律重组的双级滑阀真空泵故障诊断算法研究  

Novel Fault-Diagnosis Algorithms of Two-Stage Rotary Piston Vacuum Pump

作　　者：吕冬梅[1] 韩江[1] 于振华[1] 

机构地区：[1]合肥工业大学机械与汽车工程学院,合肥230009

出　　处：《真空科学与技术学报》2015年第5期533-537,共5页Chinese Journal of Vacuum Science and Technology

基　　金：国家重大科技专项(2012ZX04001-021);国家自然科学基金项目(51275147);安徽省自然科学基金项目(KJ2013B069)

摘　　要：为提高双级滑阀真空泵工作可靠性,针对双级滑阀式机械真空泵在实际生产过程中可能出现的故障现象,引入控制律重组和PARD-BP神经网络故障诊断算法,以2H-150型双级滑阀真空泵振动故障样本采集数据作为输入,以故障模式矩阵作为目标输出,对PARD-BP算法所得数据进行训练优化;再提出动态数据信息提取概念,进行其动态信息收集提取,采用目前广泛应用的Poly-Max模态参数识别方法进行结果验证,结果表明真空泵X、Y和Z方向上最大的峰值分别下降了31.93%,21.54%和19.37%,证实滑阀真空泵的故障诊断方法具有可靠性。Novel fault-diagnosis algorithms were developed on the basis of control law reconfiguration and PARD-BP neural network to improve the reliability of the two-stage rotary piston vacuum pump installed in industrial production line. First, the optimized training of the data, collected in PARD-BP algorithm, was conducted with the sampled vibration faults of 2H-150 two-stage rotary piston vacuum pump as the input and with the fault-mode matrix as the target output. Second, dynamic data information extraction concept was proposed to acquire the dynamic information. Finally,the opti- mized results were verified in the widely used Poly-Max modal parameter identification method. The novel algorithms were tested in experimental dynamic simulation. The test results show that the optimization resulted in reductions of the largest relative-displacement peaks in X, Y and Z directions of the vacuum pump by 31.93 %, 21.54% and 19.37 %, respectively. We suggest that the fault diagnosis algorithms be of much technological interest.

关 键 词：控制律重组 信息提取 双极滑阀真空泵 PARD-BP算法 数据训练 故障诊断 

分 类 号：TH36[机械工程—机械制造及自动化]