An automated diagnosis methodology for manufacturing and assembly failures of refrigerator compressors on production line using vibration data

This paper presents, the design of a system that performs automated detection and diagnosis of several types of failure with time-frequency (Fast Fourier Transform) based analysis of vibration data taken from hermetic compressors used in home appliances. The objective of this paper is to investigate the vibration characteristic of noisy appliances and cooling faults. Experimental studies were conducted for finding the correlation between failures and vibration data. Vibration data of appliances for the various fault cases were collected and correlation was determined between the inputs and outputs with the help of regression analysis. The fault characteristic for the huge amount of time-frequency data was transferred into a central data lake on a cloud for fault classification. Thus, characteristic algorithm was applied to automate the production process. As a result of the study, it has been revealed that the noise level of the compressors used in refrigerators with 74.41% reliability can be calculated through the vibration sensor.