Design Study of Dovetail Geometries of Turbine Blades Using Abaqus and Isight
Blades in gas and steam turbines continually face more challenging requirements for high reliability and efficiency. In order to meet these challenges in an increasingly competitive marketplace, blade design engineers are always looking for more efficient ways to design the blades in the shortest possible time and at the lowest possible cost while meeting multiple design objectives. In this paper, several design studies are performed using Abaqus and Isight to optimize the minimum contact pressure and stress around the dovetail of a typical turbine blade in order to achieved desired goals for stress levels. First, nine design parameters describing the dimensions of the dovetail are set up in a Python script which can be executed in Abaqus/CAE. The Python script generates the entire finite element model including boundary and loading conditions in Abaqus/CAE. A nonlinear static analysis considering centrifugal loading is performed in this work. After setting up the workflow using the Python script and Abaqus/CAE, Isight is used to automate the process to achieve the optimized dimensions of the dovetail. The optimization is performed in two steps. First, a surrogate model using the Optimal Latin Hypercube approximation method is created using tools in Isight. In this step, the surrogate model is used to determine the optimum values of the design variables, as well as the sensitivity of the design to the selected design variables. It also can be observed that the design is especially sensitive to five of the design variables. In the second step of the optimization, the five design variables to which the design is most sensitive are selected for further optimization by setting the other design variables to the optimized values obtained in the first step of the optimization. In this second step, several different optimization methods supported in Isight are used, including the NSGA-II non-dominated sorting genetic algorithm, Downhill Simplex, and an evolutionary optimization algorithm. Results from these methods are compared with those obtained using other common optimization methods in Isight.