Designing the predictive control of a drum dryer using multi-agent technology

Purpose. To increase the efficiency of drying lines for bulk products by automating control using intelligent technology to determine the state of the product and predict its initial moisture content by analyzing a series of control signals and messages in the time domain. Methodology. The author’s model of a drum dryer with axial and furnace burners for drying charge used for the production of iron ore concentrate – pellets – was used. The model was used to generate training and control examples. The performance of the multi-agent technology and the accuracy of predicting the initial moisture content were researched. Findings. The article analyses the factors that complicate the high-quality automatic control of the process of drying bulk products in drum dryers rotating in a furnace with burners. A model of an intelligent predictor is proposed, which identifies the state of the product and predicts its output moisture content on the basis of available control and feedback signals. The operability of the multi-agent system model and of the calculating algorithms for the predicted moisture value was proved. The possibility of using the technology to ensure automatic control of the technological process and high-quality stabilization of the controlled parameter are demonstrated. Originality. The predictor is implemented as a peer-to-peer multi-agent system. This multi-agent system stores and works signal vectors with values placed by the time delays between the change in the corresponding signal and the change in product moisture at the dryer outlet. Each agent contains a description of a specific situation in the dynamics. The technology provides for automatic adjustment of the multi-agent system by analyzing arrays of signals over a long time period and generating new agents in cases where a situation is detected which cannot be described by an array of existing agents. Practical value. The technology provides the initial moisture content calculation by an array of agents and allows the dryer automatic control by levelling the time delay in the feedback channel.