Detection of defective embedded bearings by sound analysis: a machine learning approach

Abstract

This paper describes a machine learning solution for the detection of defective embedded bearings in home appliances by sound analysis. The bearings are installed deep into the home appliances at the beginning of the production process and cannot be physically accessed once they are fully assembled. Before a home appliance is put to sale, it is turned on and passed through a sound-based sensor that produces an acoustic signal. Home appliances with defective embedded bearings are detected by analyzing such signals. The approached task is very challenging, mainly because there is a small number of sample signals and the noise level in the measurements is quite high. In fact, it is showed that the signal-to-noise ratio is high enough to mask important components when applying traditional Fourier decomposition techniques. Hence, a different approach is needed. Experimental results are reported on both laboratory and production line signals. Despite the difficulty of the task, these results are encouraging. Several classification methods were evaluated and most of them achieved acceptable performance. An interesting finding is that, among the classifiers that showed better performance, some methods are highly intuitive and easy to implement. These methods are generally preferred in industry. The proposed solution is being implemented by the company which motivated this study.

Bolt loosening detection based on audio classification

“…Machine learning detection technology based on audio classification has been widely used in the fields of medicine (Collado-Villaverde et al, 2017; Palaniappan et al, 2013), electrical fault detection (Saucedo-Espinosa et al, 2017), and speech analysis (Shin et al, 2010). However, it has not been sufficiently developed in the field of structural health monitoring (SHM).…”

Time–Frequency Envelope Analysis for Fault Detection of Rotating Machinery Signals with Impulsive Noise

“…Smith et al [17] proposed an optimized spectral kurtosis method that can be applied to signals with poor kurtogram and protrugram performance due to electromagnetic interference noise. Indeed, on the factory floor, efforts have been made to improve the detection rate of noise-defects by making soundproof rooms that block various noise, including impulsive noise [5,[29][30][31]. However, it is difficult to completely block impulsive noise because it has high energy in a wide frequency range within a short time.…”

Bearing fault diagnosis via kernel matrix construction based support vector machine

“…Rolling element bearing (REB) is a critical unit in rotating machinery and its health condition is often monitored to identify incipient fault. When a defect like bump, dent or crack that occurs in REB' outer race, inner race, roller or cage, continuously contacts another part of bearing under operation, a sequence of impulsive responses can be acquired in the form of vibration [1][2][3], acoustic emission [4], temperature, motor current, ultrasound [5], etc. However, the measured signals involve both fault-induced component and noises from structure vibration, environment interference, etc.…”