(742e) Polymer Die Design Using CFD-Based Optimization

The design of polymer dies and adapter blocks has often been more of an art than a science. It frequently requires running trial and error experiments of numerous die geometries in order to get the desired processing results. In more recent years, computational fluid dynamics (CFD) has been used to predict flow, pressure, and temperature profiles to aid the design. However, even this method normally involves a trial and error procedure to hone in on a more or less optimal design. Since there are often numerous geometric parameters and conflicting design goals, this can be quite time and compute resource intensive. With the more recent development of various optimizer software packages and the evolution of more powerful compute resources, CFD-based optimization is now practical. More time and effort may be required upfront to set up both the CFD and optimization algorithms, but then the optimizer can take over to automate the simulation runs and provide a more optimal solution in less overall time. A parameterized geometry and meshing algorithm needs to be developed to automatically create high quality meshes over the allowed geometric parameter ranges. The CFD equations, boundary conditions, physical properties and solution methods need to be set up beforehand with appropriate parameterized inputs and outputs. The geometry, mesher, and CFD software need to communicate with the optimizer software (e.g., via batch scripts). The objective functions and constraints on both input and output variables need to be defined, along with the optimization methodology. Once the algorithms are set up and linked through the optimization package, the optimizer will automatically run the CFD cases and drive towards the optimal solution without significant user intervention. The user can and should monitor the progress and can make appropriate adjustments along the way. In this paper, we will present the results of a test case to optimize a polymer die and adapter block using Gambit (ANSYS, Inc) for the geometry and mesh creation, Fluent (Ansys, Inc) as the CFD solver, and modeFRONTIER® (Esteco North America Inc.) as the optimization package. The test case involved a dogbone-shaped insert used to straighten the flow profile through the die and included 7 adjustable geometric parameters. As expected, developing a geometry and meshing algorithm that produced good meshes over the entire parameter space was one of the most challenging aspects. The main output parameters were the overall pressure drop and a measure of the flow uniformity at the die outlet. A shear thinning Carreau rheological model was used and the flow was assumed to be laminar. We first ran a single objective function optimization to minimize the standard deviation of the outlet flow, while keeping the pressure drop within specified bounds. We then added a second objective function to minimize the pressure drop as well (while staying within the specified limits.) We also explored Response Surface Model optimization methods. Since this was our first attempt at this type of CFD-based optimization modeling, we spent much of our time exploring the various optimization steps, methods, and analysis tools. Advantages and disadvantages of these methods and tools were noted. Overall, this was a very successful and insightful trial run to combine optimization software with a CFD solver to generate optimal solutions. The modeFRONTIER® software is easy to use and very powerful in terms of the number of optimization algorithms available and the analysis toolset. The parameter space can be expanded to optimize not only equipment geometry, but also processing conditions and material properties to meet various operational objectives.