Rapid analysis of coal ash composition and ash fusion temperatures using laser-induced breakdown spectroscopy

Conventional methods for coal ash composition analysis and ash fusion temperature (AFTs) determination are time-consuming, costly, and prone to subjective biases, precluding their use in real-time industrial scenarios. To address the aforementioned problems, this paper proposes a rapid analysis method for coal ash composition and ash fusion temperatures (AFTs) based on laser-induced breakdown spectroscopy (LIBS). A combination of the Multivariate Linear Regression (MLR) model and Partial Least Squares Regression (PLSR) model was employed for the rapid and accurate prediction of 10 coal ash components. This approach effectively mitigated the large prediction errors caused by the matrix effect and inter-element spectral line interference, thereby improving the prediction accuracy. For the 12 validation samples, the coefficient of determination (R2) of the 10 components ranged from 0.957 to 0.983, the Root Mean Square Error (RMSE) from 0.003% to 0.128%, and the Mean Relative Error (MRE) from 0.002% to 0.029%. Based on the contents of the 10 components, an eXtreme Gradient Boosting (XGBoost) model was constructed for the rapid prediction of AFTs. For the aforementioned 12 validation samples, the maximum absolute error of the predicted AFTs was 32.0 °C, which was significantly superior to the specification of the Chinese standard GB/T 219-2008 and satisfied the demands of practical industrial applications. The proposed method achieves rapid and accurate analysis of coal ash composition, and subsequently enables the rapid and accurate determination of AFTs, thereby providing an efficient and reliable solution for the real-time analysis of coal ash in industrial production processes.