Genetic Algorithm for Automated X-Ray Diffraction Full-Profile Analysis of Electrolyte Composition on Aluminium Smelters

Abstract

Aluminium is produced by means of the electrolysis of alumina in molten fluoride salts . A certain proportion of the fluoride compounds continuously evaporates, and this negatively impacts on the optimal composition of the electrolyte in the electrolytic baths. It means that a regular adjustment of the electrolyte composition is required by the addition of fluorides based on the results of the automatic express analysis of the electrolyte. The XRD phase analysis of crystallized electrolyte samples automatically performs the control of the main composition characteristics. This method, most frequently implemented in conjunction with aluminium smelters, necessitates periodic calibration with reference samples, whose phase composition is known exactly. The preparation of such samples is relatively complex since samples include 5–6 different phases with variable microcrystalline structure. One further diffraction method is the Rietveld method, which can be implemented without the use of reference samples. The method is based on the modelling of the experimental powder patterns of crystalline samples as the sum of the powder patterns of comprised phases, calculated from their atomic crystal structure. Included in the simulation is a refinement of the profile parameters and crystal structure of phases using the nonlinear least squares method. The difficulty associated with the automation of this approach is that a set of initial values for the parameters must be inputted that must be automatically refined by LSM to exact values. In order to resolve this problem, an optimization method was put forward by the article based on an evolutionary choice of initial values of profile and structural parameters using a genetic algorithm. The criterion of the evolution is the minimization of the profile R-factor, which represents the weighted discrepancy between the experimental and model powder patterns of the electrolyte sample. It is established that this approach achieves the required level of accuracy and complete automation of the electrolyte composition control.