Neural network will take control of Russian metallurgical furnaces
Anton Glushchenko, a Russian researcher from the Stary Oskol branch of NUST MISiS, has developed a “neural network tuner” that will increase the energy efficiency of high-power (up to 100 MW) metallurgical heating furnaces up to 10%. Material about domestic development was published in the journal "Procedia Computer Science".
The essence of innovation is that the parameters of the furnaces will be controlled by a neural network. Since the operation of blast furnaces is influenced by various factors such as burner contamination, loading and unloading of raw materials and opening curtains, the use of linear regulators with constant parameters cannot boast of high energy efficiency.
However, the “neural network tuner” proposed by Anton Glushchenko will adapt the settings of the linear regulator in real time, relying on a “bundle” of two intelligent of technologies - neural networks and knowledge bases. In this case, the first will determine the current parameters of the furnace controller and learn directly in the process.
The invention is a functional unit, which is located in the RAM of logical controllers, which are widely used in metallurgy. Its installation does not require any additional costs, since the existing control system of the unit does not need any hardware or software changes.
The essence of innovation is that the parameters of the furnaces will be controlled by a neural network. Since the operation of blast furnaces is influenced by various factors such as burner contamination, loading and unloading of raw materials and opening curtains, the use of linear regulators with constant parameters cannot boast of high energy efficiency.
However, the “neural network tuner” proposed by Anton Glushchenko will adapt the settings of the linear regulator in real time, relying on a “bundle” of two intelligent of technologies - neural networks and knowledge bases. In this case, the first will determine the current parameters of the furnace controller and learn directly in the process.
The invention is a functional unit, which is located in the RAM of logical controllers, which are widely used in metallurgy. Its installation does not require any additional costs, since the existing control system of the unit does not need any hardware or software changes.
Information