摘要
针对传统故障信号检测存在检测误差率较高、故障检测率较低等问题,提出石油工程机械设备早期振动故障信号智能检测方法。利用傅里叶变换获取设备信号中的非平稳信号,对其进行短时傅里叶变换处理,得到早期非平稳振动故障信号功率谱。在振动故障信号频域空间内进行偏最小二乘分解,依据分解结果获取潜变量,并通过显著性潜变量建立设备早期振动故障信号检测模型,实现智能检测故障信号,采用马氏距离计算功率谱中故障信号,将原始空间信号转换成相对空间信号。运用转换后的故障信号判断石油工程机械设备有无故障。仿真实验结果表明,所提检测方法与传统检测方法相比较,检测误差率降低至1.8%,故障检测率提高至99.5%。
Aiming at the high error rate and low fault detection rate in traditional fault signal detection, an intelligent detection method for early vibration fault signal of petroleum engineering machinery is proposed. The non-stationary signal of the equipment signal is obtained by Fourier transform, and the power spectrum of the early non-stationary vibration fault signal is obtained from processing the non-stationary signal with short-time Fourier transform. Partial least squares (PLS) decomposition is performed in the frequency domain space of vibration fault signals. According to the decomposition results it obtains the latent variables. The detection model of early vibration fault signals of equipment is established through the significant latent variables. The intelligent detection of fault signals is realized. The fault signals in power spectrum are calculated by Mahalanobis distance, and the original spatial signals are converted into relative spatial signals. Finally, according to the fault signal detection model constructed by statistics, the fault of petroleum engineering machinery is judged. The simulation results show that the error rate of the proposed detection method can be as low as 1.8% and the detection rate can be as high as 99.5% compared with the traditional detection method.
作者
李哲
刘宇闲
鲁晓斌
彭章保
Li Zhe;Liu Yuxian;Lu Xiaobin;Peng Zhangbao(Sixth Oil Extraction Plant of Petro China Changqing Oilfield Company, Shaanxi Xi'an, 710012, China)
出处
《机械设计与制造工程》
2019年第4期52-55,共4页
Machine Design and Manufacturing Engineering
关键词
功率谱
马氏距离
偏最小二乘
故障检测模型
power spectrum
Mahalanobis distance
partial least squares
fault detection model
