Finding the optimal solution to a problem is an every day struggle. The mankind has been in search of optimality to any problem in every aspect of life such as work. These optimization problems can differ in many aspects. For example in research work we want to find the optimal control for distillation columns, heat exchangers or membrane processes to maximize the profit and the minimize the production costs. In normal life we can also face many optimization problems without realizing it. These can be for example which way should I choose to come to work in minimum time. To solve all these optimization problems we can use several optimization methods from the field of dynamic optimization. In this research project we will focus mainly on the chemical processes.
We have to perform the following steps to solve a general optimization problem using dynamic optimization: (i) modeling step, where the process is described by mathematical equations to identify its behavior, (ii) problem formulation, where the objective of the optimization, all constraints, and decision variables are defined, (iii) the last step is the computational step, where the optimal profiles of decision variables are determined. It is important to point out that these optimal profiles only guarantee optimality for the mathematical model used. However, when applied to a real process, these profiles are optimal only when the model perfectly matches the process which is rarely the case in practice. Indeed, the uncertainty resulting from model mismatch and process disturbances are such that the model-based optimal profiles will probably not be optimal for the process. The application of these profiles to the process generally leads to the violation of certain constraints and/or to suboptimal performance.
To solve these problems several methods or techniques can be used. Optimal design experiment is one of the most widely used methods by dealing with control of real processes. The main point of this method is basically what experiments should be performed to obtain such data that they are suitable to perform parameter estimation problem for obtaining optimal values of the parameters. Once optimal parameters are identified optimal control strategy is calculated to obtain the control profiles. The last step is then to verify the simulation results with the experimental on a real process. Another interesting method from dynamic optimization methods is real-time optimization. The general idea of this approach is to use the process experimental measurements in order to improve the decision variables profiles such that the optimality conditions are satisfied by the process. In fact, for a constrained optimization problem, optimality conditions consist of two parts: feasibility and sensitivity. These two parts require different types of experimental measurements, i.e. performance index and constraints, and their corresponding gradients with respect to decision variables.
We propose the following project time schedule:
2016 (1st year)
2017 (2nd year)