１．Overview

Anomalies in mechanical equipment can be detected at a relatively early stage by monitoring vibrations rather than temperature or current. The technology for diagnosing devices through vibration measurement, applicable across many devices, enables non-destructive condition monitoring. Signal processing in the spatial domain, such as Lissajous or orbit curve analysis, is particularly effective for detecting device anomalies and failures; however, it involves challenges such as data synchronization, noise interference, and complex sensor adjustments.

We have developed a new technology for easier and more accurate three-dimensional vibration data processing, encompassing both vibration sensors and analysis methods. The developed sensors are easy to install, and their combination with a preprocessing algorithm for precise Lissajous curve drawing enables accurate grasp of the vibration state of the equipment.

２．Development Overview and Application Examples

A three-axis digital vibration sensor optimized for Lissajous curve drawing are developed (Fig.1), boasting low noise, high resolution, and superior synchronization. This sensor utilizes a “quartz based double ended tuning fork transducer (Fig.2),” whose resonant frequency changes according to acceleration, and realizes high sensitivity and synchronized measurements through its combination with subsequent digital processing.

On the software side, a preprocessing algorithm was developed to lower background noise and extract signals aligned with mechanical vibrations. By concentrating on the vibration cycle to segment and average data, we efficiently isolated essential signals, maintaining phase information across axes for more precise analysis than traditional methods.

 　　  
(a) Water/dust  proof model     (b)  Build-in model

Fig.1 Sensor appearance

Fig.2 Quartz transducer

For application, we tested on two identical electric motors, A and B, installed in the same year, and operated over 10 years but with different operational histories; Motor A was overhauled two years ago, Motor B ran continuously for ten years. Although time and frequency analysis showed no clear differences, Lissajous curves revealed distinct vibration patterns (Figs.3 and 4): Motor A’s symmetrical ellipse versus Motor B’s distorted shape, indicating how operation time influences vibration states. This method also allows for quantitative analysis of these patterns.

３．Summary and Future Outlook

Three-axis digital vibration sensors with low noise and high precision are developed, alongside a diagnostic technology that simplifies signal processing in the spatial domain. These sensors are easy to install and can effectively track changes in the operating state of machinery over time, which is expected to improve maintenance efficiency in factories and infrastructure.

These sensors are implemented in systems for monitoring the condition of sluice gates in rivers and dams, aiming to pinpoint correlations between deterioration and device anomalies. Future development efforts will focus on refining the use of Lissajous curve drawings as a viable diagnostic tool, through optimal indexing and the introduction of machine learning.

Fig.3 Electric motor with sensor installed

(a)Motor A for 2 years　　(b)Motor B for 10 years

Fig.4 Difference in Lissajous curves

Kenta Sato(Seiko Epson Corporation)
Masayoshi Todorokihara(Seiko Epson Corporation)
Masayuki Oto(Seiko Epson Corporation)
Toshio Takiya(Hitachi Zosen Corporation)
Takaharu Kitamura(Hitachi Zosen Corporation)