Quantitative analysis of the major components of coal ash using laser induced breakdown spectroscopy coupled with a wavelet neural network (WNN)

A laser induced breakdown spectroscopy (LIBS) technique was applied to detect the major components of coal ash based on a wavelet neural network (WNN). Prior to constructing the WNN model, the spectra were preprocessed using wavelet threshold de-noising and Kalman filtering, and the principle components (PC), extracted using principle component analysis (PCA), were used as the input variables. Afterwards, the quantitative analysis of the major components in coal ash samples was completed using the WNN with the optimized WNN model parameters consisting of the number of hidden neurons (NHN), the number of iterations (NI), the learning rate (LR) and the momentum based on the root mean square error (RMSE). Finally, an artificial neural network (ANN) and the WNN were evaluated comparatively on their ability to predict the content of major components of test coal ash samples in terms of correlation coefficient (R) and RMSE, demonstrating that LIBS combined with a WNN model exhibited better prediction for coal ash, and is a promising technique for combustion process control even in the online mode.