This article reports the thermal ageing of ester fluid-impregnated pressboard material along with its performance evaluation using optical emission spectroscopy(OES)and classifi-cation using machine-learning algorithm...This article reports the thermal ageing of ester fluid-impregnated pressboard material along with its performance evaluation using optical emission spectroscopy(OES)and classifi-cation using machine-learning algorithms adopting laser-induced breakdown spectroscopy(LIBS).The surface discharge analysis on ester-impregnated pressboard(EIP)is studied using OES and the plasma temperature was evaluated based on the Cu I emission lines which were higher under the negative DC compared with the positive DC and AC voltages.The LIBS analysis was performed on the EIP material operated at different energy levels in order to acquire the optimal energy required to be used for its classification algorithm.The intensity ratio and electron density evaluated from LIBS studies correlated well with the plasma temperature.The lower limit of detection(LOD)calculated based on linear regression analysis for copper peak was around 3.5 times higher than the identification of carbon peak.The machine-learning techniques like principal component analysis and neural network algorithm have been performed on the LIBS spectral dataset in order to classify the ageing of EIP material.Artificial neural network adopting LIBS provided a better classi-fication accuracy on all the aged samples compared with principal component analysis which classified only for the samples aged at higher temperatures.展开更多
文摘This article reports the thermal ageing of ester fluid-impregnated pressboard material along with its performance evaluation using optical emission spectroscopy(OES)and classifi-cation using machine-learning algorithms adopting laser-induced breakdown spectroscopy(LIBS).The surface discharge analysis on ester-impregnated pressboard(EIP)is studied using OES and the plasma temperature was evaluated based on the Cu I emission lines which were higher under the negative DC compared with the positive DC and AC voltages.The LIBS analysis was performed on the EIP material operated at different energy levels in order to acquire the optimal energy required to be used for its classification algorithm.The intensity ratio and electron density evaluated from LIBS studies correlated well with the plasma temperature.The lower limit of detection(LOD)calculated based on linear regression analysis for copper peak was around 3.5 times higher than the identification of carbon peak.The machine-learning techniques like principal component analysis and neural network algorithm have been performed on the LIBS spectral dataset in order to classify the ageing of EIP material.Artificial neural network adopting LIBS provided a better classi-fication accuracy on all the aged samples compared with principal component analysis which classified only for the samples aged at higher temperatures.
