A novel strategy for quantitative analysis of the energy value of milk powder via laser-induced breakdown spectroscopy coupled with machine learning and a genetic algorithm

Abstract

The energy value of milk powder is an important indicator of its nutritional value, meaning it is of great significance to explore methods of quickly detecting this energy value. In this study, laser-induced breakdown spectroscopy (LIBS) combined with an extreme learning machine (ELM) algorithm was applied to quantitatively study the energy value of milk powder. First, a full-spectrum ELM model was established. To improve the prediction performance, a competitive adaptive reweighted sampling (CARS) algorithm was introduced to filter 4096 wavelength variables of milk powder, with 114 of them selected to build the CARS-ELM model. Second, a genetic algorithm (GA) was used to optimize the weight and bias of the ELM and CARS-ELM models, respectively. The results show that the GA-CARS-ELM model obtains the best predictive performance, with the R_P², RMSE_P and MAPE_P of GA-CARS-ELM being 0.9927, 0.2349, and 1.20%, respectively. This indicates that LIBS combined with the GA-CARS-ELM model can accurately predict the energy value of milk powder.