NB1 Demo
Continuous fermentation process
Inside a bioreactor vessel, millions of microorganisms consume a nutrient substrate
and metabolize it into a commercially valuable product. The controller's objective
is to maintain the substrate concentration at some desired optimal level by adjusting
the flowrate of growth media through the vessel. The reaction kinetics exhibit substrate inhibition
as well as product inhibition, and the dynamics are therefore complex and very nonlinear.
System equations for the continuous fermentation process
X, S, P, and Sf are
the cell, substrate, product, and feed concentrations, respectively,
D is the dilution rate, μ
is the specific growth rate, μm
is the maximum specific growth rate, and the remaining variables are yield parameters
and constants.
Tham et al.
1
developed a controller for this system using sliding mode control (SMC —
a technique that involves Lyapunov functions, manifold sliding surfaces, adaptive switching gains,
and other kinds of intricate math). They also developed a PI controller.
The following sections compare the performance of these two controllers
to that of an NB1 through several different types of disturbances.
Setpoint changes
The following figure shows the performances of the SMC controller and the PI controller
through a series of setpoint changes.
SMC and PI control
Reproduced from Fig. 8 of Tham et al.
1
The following figure shows the performance of an NB1,
which we plugged in without any tuning, on the exact same problem.
NB1 control
The NB1 tracks the setpoint with much less lag.
The following figure shows the reactor dilution rate as manipulated by the NB1 in
achieving the performance shown above.
NB1 output
Load disturbances, part I
The difference in performance between the SMC and PI controllers
is more dramatic in the following figure, which shows the response
to a series of disturbances in the feed concentration, Sf.
SMC and PI control
Reproduced from Fig. 9 of Tham et al.
1
The SMC method, which is known mostly for its
exceptionally robust control, is a clear improvement over the PI.
The NB1, as can be seen in the following figure,
is an even further improvement.
NB1 control
The following figure shows the NB1's output.
NB1 output
Load disturbances, part II
The final figures show the responses to disturbances in the
maximum specific growth rate, μm.
SMC and PI control
Reproduced from Fig. 10 of Tham et al.
1
NB1 control
NB1 output
Once again, the NB1 bests the PI controller and the advanced SMC controller.
1
H.J. Tham, K.B. Ramachandran, and M.A. Hussein.
Sliding mode control for a continuous bioreactor.
Chem. Biochem. Eng. Q. 17 (4): 267-275 (2003).