| This work deals with the use of a multipurpose simulator as a tool for process operability analysis and has focused on overall control system design. The way to handle this very large problem is to reduce it to smaller subsystems, i.e., decomposition. In order for the decomposition to be meaningful, the subsystems should have minimal interaction. Thus, some means of quantifying the interaction between subsystems must be available.; Three common measures of interaction have been compared to determine their applicability to large chemical processes: the dynamic relative gain array, singular value analysis, and the inverse Nyquist array. To test these methods, an array of unit operations models has been developed. Each model can perform three types of simulation: steady-state, dynamic, and frequency response. Unlike other simulators of this class, the latter is found directly from the linearized Laplace-transformed dynamic equations and not from dynamic testing.; The three decomposition methods have been applied to a number of distillation columns, an example of two heat-integrated columns, two heat exchanger networks, a methanol synthesis process, a methanation process, and a dimethyl ether synthesis process. The problems ranged in size from two to twelve control loops.; The dynamic relative gain array was found to be the most reliable measure of interaction because it is invariant to scaling and it gives the user both an insight into the nature of the interactions and an ability to estimate the sensitivity of the system. The inverse Nyquist array was found to give identical results to the dynamic relative gain array for steady-state, two-by-two systems when scaled so that the condition number (a measure of process sensitivity to modeling errors) was minimized.; A complete interpretation of the dynamic relative gain array has been presented. This includes not only the interpretation of the magnitude of the relative gain, which has been the main focus of previous work, but also of the phase angle which was found to be as important as the magnitude.; Finally, a methodology for overall control system design based upon the dynamic relative gain array has been presented. |
