Optimizing the process parameters in stirred-media mills using mathematical modeling

This study investigated the interactions among process parameters in stirred-media mills to achieve precise control of dissociation parameters for spodumene adaptability. Response surface methodology (RSM) and back propagation (BP) neural network models were employed to predict and optimize the three key process parameters based on the grinding technology efficiency as the evaluation criterion. The order of significance of the parameters affecting the grinding efficiency was as follows: grinding time &gt; media filling rate &gt; stirrer speed. The optimal grinding parameters were identified as a media-filling rate of 52%, stirrer speed of 506 rpm, and grinding time of 3 min. Under these conditions, the RSM model achieved an <i>R</i>² value of 0.9509, relative error of 13.15%, mean squared error (<i>MSE</i>) of 0.0011, and mean absolute error (<i>MAE</i>) of 0.0276. Conversely, the BP neural network model achieved a higher <i>R</i>² value of 0.9950, a relative error of 4.77%, <i>MSE</i> of 0.0003, and <i>MAE</i> of 0.0091. Although the RSM model effectively optimized the grinding process parameters, the BP neural network model exhibited superior fitting performance and accuracy in predicting the grinding efficiency under varying process parameters. This study provides valuable technical insights for optimizing process parameters in the grinding dissociation process.