Detection Method for Flotation Tailings Ash Content Based on Regularized Stochastic Configuration Networks Dynamic Compensation

The ash content of flotation tailings is a crucial parameter in flotation operations. This paper proposes a ash content predicting method via capturing infrared backscattered light from flotation tailings with multiple light scattering instruments in response to the current challenges of online ash content detection technologies, such as high costs, environmental pollution, and the discrepancy between the costs of detection principle technology and accuracy. The linear relationship between light intensity and concentration is established by Mie theory when the ash content is inversely proportional to light intensity. Based on this relationship, a tailings ash content prediction model which taking concentration correction into account is developed. In this study, various ash content tailings with different concentration ranging from 0 g/L to 80 g/L were selected to research the coupling effects of backscattered light intensity with ash content and concentration. The industrial experimental results show that the L-RSC model obtained by combining the linear model and nonlinear RSC model based on regularized stochastic configuration network has good performance. The model achieved a 68.75<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> of detected values within an acceptable error margin of (δ<1.0%, along with an average absolute error (MAE) of 0.54, a root mean square error (RMSE) of 0.69, a coefficient of determination (R2) of 99.21 % and a detection duration of 5 minutes. This study aims to propose an offline, passive, and cost-effective detection method for coal tailings and establish a data-driven prediction model to facilitate the detection of ash content in flotation tailings, thereby opening up new avenues in this field.