Hu Ruo, Feng Jinyang, Mou Zonglei, Yin Xunlong, Li Zhenfei, Ma Hongrong
National Institute of Metrology, Beijing 100029, China.
College of Electrical Engineering and Automation, Shandong University of Science and Technology, Qingdao 266590, China.
Rev Sci Instrum. 2022 May 1;93(5):054501. doi: 10.1063/5.0079424.
The vibration disturbance caused by incipient faults is an important factor affecting the measurement accuracy of the cam-driven absolute gravimeter. Based on the characteristics of the cam-driven absolute gravimeter, such as the small amplitude of the incipient faults, the inadequate representation of features for the faults, and hard-to-find in the noise, a novel method for incipient fault diagnosis of the cam-driven absolute gravimeter is put forward in this paper, which integrates the parameter-optimized Variational Mode Decomposition (VMD) with Light Gradient Boosting Machine (LightGBM). The sparrow search algorithm is used to optimize the VMD parameters. The parameter-optimized VMD algorithm is used to adaptively decompose the vibration signals of the gravimeter under different cases, and then an effective intrinsic mode function (IMF) is selected based on the Pearson correlation coefficient. Some high-frequency IMFs are subjected to adaptive noise reduction combined with low-frequency IMF reconstruction, and then the multi-scale permutation entropy with sensitive characteristics under different time scales is extracted as the fault feature vectors. The extracted multi-dimensional vector matrix is entered into the LightGBM classifier to realize the accurate diagnosis of the incipient faults for the cam-driven absolute gravimeter. The test results show that this method can effectively detect various incipient failures of the cam-driven absolute gravimeter, with an identification accuracy of 98.41%. With this method, the problem of low measurement accuracy for the cam-driven absolute gravimeter caused by the incipient faults is solved, and the rapid tracing and accurate positioning of these faults for the gravimeter are realized, promising a good prospect for engineering application.
早期故障引起的振动干扰是影响凸轮驱动绝对重力仪测量精度的重要因素。基于凸轮驱动绝对重力仪早期故障幅度小、故障特征表征不足以及在噪声中难以发现等特点，本文提出了一种将参数优化的变分模态分解（VMD）与轻量级梯度提升机（LightGBM）相结合的凸轮驱动绝对重力仪早期故障诊断新方法。采用麻雀搜索算法对VMD参数进行优化。利用参数优化后的VMD算法对重力仪在不同工况下的振动信号进行自适应分解，然后基于皮尔逊相关系数选择有效的本征模态函数（IMF）。对部分高频IMF进行自适应降噪并结合低频IMF重构，接着提取不同时间尺度下具有敏感特征的多尺度排列熵作为故障特征向量。将提取的多维向量矩阵输入到LightGBM分类器中，实现对凸轮驱动绝对重力仪早期故障的准确诊断。测试结果表明，该方法能够有效检测凸轮驱动绝对重力仪的各种早期故障，识别准确率为98.41%。该方法解决了早期故障导致凸轮驱动绝对重力仪测量精度低的问题，实现了重力仪这些故障的快速追踪和精确定位，具有良好的工程应用前景。
