Predicting the production efficiency of barium metaborate using machine learning

Abstract Barium metaborate (BBO) is a crucial additive that enhances the resistance of materials against detrimental effects. BBO has unique antimicrobial, radiation shielding and flame-retardant properties. The β form of barium metaborate is of great significance in producing single crystals, the development of functional glasses, and the advancement of laser technology. Previously, we reported the synthesis of barium metaborate crystals from barium chloride and sodium metaborate precursor solutions using a chemical precipitation method. In this following study, the production of barium metaborate was modelled using machine learning (ML) tools. The optimum procedure conditions were investigated using the response surface optimization (RSM) method. The study tested five different ML methods, and the random forest (RF) method giving the most successful results was used for more detailed analysis. The RF model achieved accuracy scores of 0.91 for yield1 and 0.93 for yield2, with corresponding MSE values of 12.58 and 7.33, and RMSE values of 3.55 and 2.71, respectively. In addition, the significant yield increase at 80 °C was highlighted particularly evident in its effectiveness at enhancing the production efficiency of BBO in industries. When the 49 experimental data results were examined, it was seen that the higher and more stable yields were obtained when NaOH was used as a solvent in comparison to HCl. The experimental results were found to be in close agreement with those obtained using RSM, thereby validating the findings of response surface optimization.