Experimental investigation of subspace identification and LQ control of a pressure tank

Many control studies of complex processes are limited to simulation studies. However, many real processes exhibit peculiarities and nonlinearities that are difficult to model. Also some control designs are difficult to implement online in practice. Because of the recent availability of economical real time software, more complex control algorithms will see real time testing and implementation. In spite of this, many compromises are often necessary when the final control is implemented. This paper follows the identification (using subspace methods) and real time linear quadratic control of a pressure tank system and points to some of the compromises in identification and control design that are often necessary. A multivariate control is implemented on a pressure tank process with rather good results in spite of these compromises.