Multilayer perceptron as the tool for modeling of reaction crystallization of barium sulphate in MSMPR crystallizer

One of the most ecologically harmful industrial wastes are the post-processed, used quenching salts, especially rich in BaCl₂. Original method of their neutralization is based on their dissolution in water followed by a complex reaction crystallization process (after solid (NH₄)₂SO₄ addition) effecting in production of barium sulphate. The process regime, which determines the crystalline product quality, depends on many technological parameters which individually influence various partial processes in micro- and macroscale. Facing this intrinsic complexity of the process its reliable analytical model has not been elaborated up till now. Application of artificial neural networks (e.g. multilayer perceptrons) for thorough description of such complex systems is substantiated since these need only the scattered information incorporated within the raw experimental data. An alternative model of the system behavior - numerical but free of any simplifying assumptions - is thus possible. Neural network simulation effects concerning reaction crystallization of barium sulphate in a DT MSMPR crystallizer are presented and discussed.