Multi-objective global optimization of a butterfly valve using genetic algorithms

- This paper deals with the multidisciplinary design optimization of a double eccentric butterfly valve taking into account the structural safety of the valve, its fluid dynamic influence and its weight.
- In order to obtain an optimum design in this paper 3D structural (FEM) and fluid dynamic (CFD) evaluations have been carried out simultaneously.
- The optimization process is set up by means of different Python scripts that connect the FEA, CFD and Genetic Algorithm software.
- A new parametric model based on control points in order to manage the design of the back side of the valve is proposed enabling us to explore a much wider range of solutions.
- As different confronting objectives need to be balanced and the computational time of the evaluations is considerable an efficient optimization algorithm (in this case NSGA-II) is needed.
- Selecting the genetic algorithm (NSGA-II) based on the Pareto dominance approach allows us to obtain a set of optimum designs (Pareto Front) and not only a single optimal solution.
- The optimization result shows that the GA found an optimum valve design that decreases the maximum stress by 65.4%, the mass by 5 % and increases the Kv by 11.3 %.
- The results obtained confirm that the proposed methodology in this study is adequate for addressing the optimization of these types of valves.

A butterfly valve is a type of valve typically used for isolating or regulating flow where the closing mechanism takes the form of a disc. For a long time, the attention of many researchers has focused on carrying out structural (FEM) and computational fluid dynamics (CFD) analysis in order to increase the performance of this type of flow-control device. This paper proposes a novel multi-objective approach for the design optimization of a butterfly valve using advanced genetic algorithms based on Pareto dominance. Firstly, after defining the need for this study and analyzing previous papers on the subject, the initial butterfly valve is presented and the initial fluid and structural analysis are carried out. Secondly, the optimization problem is defined and the optimization strategy is presented. The design variables are identified and a parameterization model of the valve is made. Thirdly, initial design candidates are generated by DOE and design optimization using genetic algorithms is performed. In this part of the process structural and CFD analysis are calculated for each candidate simultaneously. The optimization process involves various types of software and Python scripts are needed for their interaction and the connection of all steps. Finally, a set of optimal solutions is obtained and the optimum design that provides a 65.4% stress reduction, a 5% mass reduction and a 11.3% flow increase is selected in accordance with manufacturer preferences. Validation of the results is provided by comparing experimental test results with the values obtained for the initial design. The results demonstrate the capability and potential of the proposed methodology.