Cokes ovens produce cokes by heating coal to (approx) 1400 degrees Celsius in an oxygen deficit environment. The cokes is an essential 'ingredient' for making steel in a blast furnace (cokes is mainly used as a 'fuel'). During this process, of heating coal in cokes furnace, the coal releases poisonous gasses that are sucked off via valves and ventilators to chemical cleaning installations. At Tata Steel, PI controllers adjusted these valves and ventilators based on pressures. The most essential pressure is the pressure inside the cokes furnace. This pressure must be maintained (as well as possible) at a fixed setpoint value above the environmental pressure (just outide the furnace). If this pressure is not controlled well, and it drops too low, too much air is sucked in and the furnace walls are damaged. If the pressure rises too much, the poisonous gasses escape to the environment.
DotX was asked to check the tuning of the controllers and to investigate any other potential improvements.
We found out that the PI controller were not tuned optimally, and that control could be improved by adding D-action to the PI controller. In addition, we found out that control could be improved by extending the PID controllers with self-learning feedforward (Iterative Learning Control). We helped implement the proposed changes to the control system. As a result, the standard deviation of the pressures reduced by 40%. The figure below shows measurements BEFORE and AFTER the control changes.