Recycling high-purity rare earths from the waste in fluorescent lamps will assist in preserving rare earth resources. However, the present recovery from the waste in fluorescent lamps is very low at around 30%, and most waste fluorescent lamps are reclaimed. In this study we attempted to reduce one cause of difficulty in recycling lamps, namely the identification of rare earths in plastic-covered lamps, which is conventionally executed by human operators. Using the Acoustic Emission (AE) signal of the collision sound between lamps and an AE sensor head, we first carried out waveform analysis, and frequency analysis using the discrete Fourier transform. We extracted six feature values for each analysis and differentiated the plastic-covered lamps using thresholds. The least error rates were 4.85% and 6.67% for the waveform and frequency analyses, respectively. Second, we applied discriminant analysis on the six feature values from each analysis, resulting in error rates of 3.03% and 5.45%, respectively. Third, we extracted six feature values from the time-frequency analysis using the discrete Wavelet transform and Fractal dimensioning, and applied discriminant analysis on the six feature values, resulting in an error rate of 4.24%. Finally, we examined all the results from the three analyses to find the best combination of feature values, applied discriminant analysis to 12 feature values from the time analysis and time-frequency analysis, and succeeded in reducing the error rate to 0%.
|Translated title of the contribution
|Classification of the Waste Fluorescent Lamps using Signal Processing and Discriminant Analysis
|Number of pages
|Journal of MMIJ
|Published - 2016