Cokes furnace control improvements

Problem: reduce pressure variations in Cokes furnace 2.
Result: 70% reduction of pressure peaks

First strike (2010)
Cokes furnaces produce cokes from coal. During production, poisonous gasses escape from the coal and are sucked off and transported to the gas cleaning plant.

It is important that the pressures in the furnaces are kept close to setpoint with minimal variations. If the pressure peaks too high, it will escape to the environment. If the pressure is too low, air enters the furnace and burns the coal/cokes and damages the furnace lining. To allow precise control, the pressure at 5 locations in the furnace is controlled at setpoint by 5 PID controllers that adjust valve positions.

In the first phase, DotX analysed data and investigated the cause of the gas pressure variations and the possibilities to reduce these by adjusting the control system. In 2010, based on this analysis, the first changes were made to the control system:

1. PID tuning was optimised
2. Self Learning Feedforward was implemented as an add-on to 4 PID controllers
3. Play compensation was implemented as an extension to one of the control outputs

Self Learning Feedforward acts solely on a trigger signal, that goes to 1 if the furnace is charged with new coal. The Figure below shows the (averaged) pressure variations with Self Learning Feedforward On (left) and OFF (right). Clearly, the pressure peaks are reduced considerably.
The play compensation reduces the effect of play between the control output (demanded valve position) and actual valve position. All in all, the pressure variations, decreased by 50%.



Second strike (2020- now)
In 2020, Tata asked advice on how to reduce the pressure variations further. We then came up with the following possibilities:

1. Implement real-time monitoring of all PID loops (pressures controlled, outputs of PID, and valve position signal)
2. Make model of the pressure using first principles, verify it, and use the model to simulate a variety of control system improvements

The first item, real-time monitoring, has been implemented by DotX in the Azure environment in 2021. This web access allows only to monitor data, and does not provide access to any Tata computer or PLC. The monitor allows (those with access) to check loop health, and PID inputs and outputs at any time and anywhere.

The figure below shows a screenshot of my mobile phone (where the internet address has been changed for security reasons). It shows the Process Variable (PV) and Setpoint (SP) during the last hour. If a loop suffers an issue, a group of people (including DotX ) is send an email, explaining the issue, and asking to take action. This email alert is additional to the regular alarms in the operator room (which are sometimes missed), and it provides us (DotX) the opportunity to react.



Every month, based on the data, DotX creates a report on Key Performance Indicators (KPIs) and discusses the outcome with Tata. The KPIs include standard deviations of PV-SP, a frequency spectrum analysis (to detect badly damped oscillations), estimations of play in each actuator, and the effectiveness of the control system during charging (where the highest pressure peaks occur).

The second item (modelling and analysis study) was executed in 2020. It was concluded that adding Gain Scheduling, and making several changes to feedforward were promising measures.

Each change to the control system was evaluated using proven scientific procedures, like switching every minute between a change and the old situation for at least 5000 times. This allows to compute the effect on all PID loops quite accurately. If a measure seems successful, it is monitored over a longer period of time to check its ‘health’. Some of the adjustments required additional analog communications between PLCs which were realized by Tata.

Results and lookout
Due to the continuous efforts in improving control, the standard deviations of each loop are likely to continue to reduce, as they have done so far.

Back to Control Projects