45400-AC9
Interval Operability as a Design Tool for Model Predictive Controllers
During the second year the project was active only for half of the year due to the graduation of the first PhD student and the initiation of a new student in the spring semester of 2008. The effort in 2008 has been focused on the extension of the previous results into the area of nonlinear high dimensional problems. In this type of problems the task of calculating in detail the Achievable Output Set (AOS) of an n-dimensional Available Input Set (AIS) through a nonlinear model is an impossible task computationally.
To overcome this obstacle, we are proceeding in calculating these related sets by an approximate method. I this new approximate calculation we are selecting a set of specific points in the input set for which we calculate the corresponding points in the response set. This calculation is quite doable computationally. However of critical importance is the systematic selection of the number and the specific location of such input points so that so the polynomial interpolation can provide an accurate enough approximation on the calculated points in the output space. To solve this problem we have started utilizing techniques used in the design of experiments. Here the “experimental process” is the complex process simulator.
Initially we successfully addressed problems of medium dimensionality. We are presently aiming to test our ideas in the rather complex process simulation known in the literature as the Tennessee Eastman Process.
