COMPUTER-ASSISTED GRAPHICAL OPTIMIZATION OF CHEMICAL PROCESSES

Abstract

In this thesis a new graphical optimization method has been developed. This method is suitable for solving global optimization since the graphical information of the terrain described by the function allows the use of the visual heuristic search to locate the most promising region of the global optimum. The advantage of the graphical method over numerical methods is that the former can be used as a general optimization solver which can overcome problems such as convergence to a local minima, premature termination by non-smooth terrain and discontinuity. This method uses the novel graphical format called HGRAM graphical format in association w.th the axiom and theorems of Point Geometry. This format allows one to map n-dimensional points onto two-dimensional display media. In this method, the function and its independent variables in the Rn are mapped bijectively as a set of points on layered manifolds the trajectories of which are generated using a set of simple parametric functions which are stepped up to vary their values. The search of the global minimum is sped up by using an interpolation method which approximates the minima locus path along the circular path fitted by three points. The graphical method is found to perform very well in several tests using standard functions in term of accuracy attained compared to the results by numerical methods. Two examples of classical problems in chemical engineering, i.e. yield optimization of CSTRs in series and process design optimization of Williams & Otto plant, are tested. The results obtained are comparable to those obtained by previous workers minus the algebraic complexity.