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

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