Low-sampling rate data-based failure diagnosis by using self-powered system

Abstract

In recent years, in order to solve critical problems such as global warming and climate change that have been occurring in the world, research on the application of power generation methods which are harmless to the environment has been underway. In this paper, a fault diagnosis method using the self-powered sensing system based on vibration power generation is proposed. Frequency analysis is known as a general failure diagnosis method. However, due to the limitation of the generated power, the sampling period of data acquisition is as large as a few-ten millisecond. For this reason, it is difficult to use frequency analysis for fault diagnosis using vibration power generation. Therefore, a fault diagnosis system corresponding to an increase in the sampling period is constructed by introducing machine learning. An acceleration sensor used for data acquisition is driven by the vibration power generator attached to factory equipment. The diagnosis is performed by wireless-transmitted acceleration data. By introducing a machine learning strategy into the diagnosis, accurate diagnosis can be performed even for data with low-sampling rate. The effectiveness of the proposed diagnosis method is experimentally evaluated by using the factory equipment.