Figure 1: screenshot of the PID Tuner 'Fit Model' section showing the result of the step experiment (step on the coolant valve position)
Figure 2 shows the complete log. The first half hour was used for step/identification experiments. After half an hour we had gathered sufficient data to identify the process model. Immidiately after that, we tuned the PID controllers and switched them on. The temperature loop is relatively fast compared to the density loop, but both loops reduced the process variable variations with more than 80% (compared to manual control). Therefore, after approximately 1 hour, the operator agreed to increase the density setpoint, and to reduce the temperature setpoint. As a result, the heat valve position changed from 60% (at time 0) to 55% (at the end of the experiment) which boils down to approximately 8% reduction of steam flow. The automatic control system has been in operation ever since the tuning experiment.
Figure 2: complete log of the experiments. The first half hour was used for step experiments. Based on the results of those experiments, the PID controllers were tuned and switched on. After 2.2 hours the temperature setpoint was reduced resulting in lower steam flow consumption (lower valve position means less steam flow consumption).