Tata steel


Cokes furnaces are heating coal at approximately 1400 degrees Celcius to produce cokes, as an ingredient in ironmaking. When heating coal, poisenous gasses escape from the coal. These gasses are transported to chemical 'cleaning' installations via ducts with ventilators and valves. PID controllers adjust the ventilators and valves to control the pressure in the furnace at a fixed level (just a couple of millibars) below the air pressure outside the furnace. If the furnace pressure drops too low, too much air enters the furnace and 'burns' the cokes and causes damage to the furnace. If the pressure is too high, the poisonous cokes gasses escape into the environment. Hence, it is important that the PID controllers in the furnaces are tuned optimally, and to know if there are more options to improve control (i.e. reduce pressure variations).

In 2010, the PI controllers were extended with D-action, and the PID controllers were tuned optimally. Furthermore, Learning Feedforward was added. As a result, the pressure peak errors were reduced by 50%, and the standard deviations of the gas pressure variations were reduced by 40%. The figure below shows measurements of the pressure in blue (scaled) BEFORE (i.e. without learning FF), and AFTER (with Learning Feedforward).

Result: 70% reduction of pressure peaks