An incremental optimizing controller for constrained nonlinear processes

Chemical process control today, has its own share of problems and deficiencies and there is plenty of room for improvement. These problems and issues include nonlinearity, multivariable processes, interaction and coupling, disturbance rejection, constraint handling, computational speed, ease of design and tuning, degrees of freedom, initialization and transfers, and ill-behaved dynamics. The conventional PI/PID controller is still largely used in industrial control loops today. But, it is essentially a linear controller and requires a lot of tack-on fixes to cope with the above mentioned problems. This makes its design complex and the effort required quite laborious. Also, advanced control algorithms like Model Predictive Control entail computational complexity when it comes to dealing with process nonlinearity.

In this work, a novel nonlinear control approach called Incremental Optimizing Control (IOC), has been developed. The key embodiment of this control approach is the prediction of process gains and the incremental changes based on it. Gain prediction is done using neural network models that facilitate rapid and efficient computation. Constraints and degrees of freedom issues are solved by optimization. This approach has the potential to tackle all of the problems/issues mentioned in the previous paragraph with the exception of ill-behaved dynamics. Since for a lot of chemical processes, nonlinearity rather than ill-behaved dynamics is the dominant problem, this approach should therefore have its own significant applications. Besides coping with nonlinearity, IOC has the ability to handle constraints, take care of interaction effects, provide good disturbance rejection, be simplistic in design and be computationally less expensive.

The IOC approach has been developed for Single-Input Single-Output (SISO) and Multiple-Input Multiple-Output (MIMO) processes, starting from basic principles. Theoretical analyses place this approach on firm fundamental ground. Excellent results have been obtained from the use of IOC on a nonlinear, multivariable, multicomponent flash process. These include very good setpoint tracking and disturbance rejection, effective decoupling and excellent constraint handling.