A novel approach to the control of dynamically unfavorable processes

It has been suggested by Rosenbrock [8] that a method to improve the over-all control of a complex process like a distillation column is to use a theory based on matrix algebra to adjust the dynamic response of the process to some required response before applying conventional control. The development has the main drawback that a very large number of control and measurement variables are required. Practical considerations require that these be reduced. The method suggested is quite generally applicable to any linear multivariable system. It is shown that, when only a limited number of variables are used for control and measurement, the maximum improvement possible can be estimated from asymptotic considerations. Further this limit on system performance is dependent only on the process and the number of variables available for control and measurement, not on the control system details. The theory was applied to a linearized model of a five-plate binary distillation column and the lowest natural frequency was increased by a factor of five. This paper includes neither details of how this system of control would be realized practically nor what is the desired response of the system before applying conventional control. This is a preliminary survey of the prospects for this system of control and with the encouraging answers obtained it will be worthwhile pursuing the method further.