Optimal design of oil and gas industry equipment through enumeration techniques

Abstract

A significant portion of the costs associated with the oil and gas industry corresponds to the process equipment. Therefore, many researchers have been devoting effort to develop new approaches to solve the optimal design problem in order to reduce costs. Two important classes of methods are extensively investigated in the literature for the solution of design problems: metaheuristic methods and mathematical programming. However, despite their success to solve several design problems, both approaches present some relevant limitations (e.g. mathematical programming algorithms may not converge, the performance of metaheuristic optimization depends on an adequate tuning of the control parameters, etc.). Aiming at proposing an optimal design procedure without the limitations of previous approaches, this dissertation explores the utilization of two enumeration techniques: Partial Set Trimming followed by Smart Enumeration and SuperSmart Enumeration. Partial Set Trimming employs the inequality constraints to reduce the number of solution candidates. Smart Enumeration involves the enumeration of the set of solution candidates according to a crescent order of a lower bound of the objective function, associated with a proper stopping criterion based on the value of the incumbent objective function. The SuperSmart Enumeration is a novel procedure, proposed in this dissertation, which accelerates the solution of design problems using candidate elimination rules (Similarity Elimination and Elimination by Objective Function). These rules are aggregated in a recursive search from top to bottom of a set of solution candidates organized in a crescent order of their dimension. Both enumeration procedures do not present convergence problems and always attain the global optimum. Their performances were tested here for the design of gasketed-plate heat exchangers and fixed-bed reactors. A comparison with metaheuristic methods showed that the enumeration procedures lead to lower computational times and the SuperSmart Enumeration provides a large reduction of the computational time when compared with the Smart Enumeration.