The paper describes the application of an optimization method to the redesign of a turbocharger compressor wheel. The starting design presents quite high performance. Previous attempts to improve this design have shown it difficult to increase aerodynamic performance without compromising mechanical stress levels. The optimization methodology relies on the combination of a genetic algorithm, a neural network, a database, and user generated objective functions. The originality of the paper is that the optimization is not only coupled to a CFD solver, but also to a CSM solver, so that mechanical stresses can be included in the optimization objectives. A parametric model of the solid sector of the blade, back plate and bore zone is built and included in the optimization. The challenging turbocharger test case has allowed gaining experience with design objectives of different nature. The results show that the optimization has been able to improve the aero performance, while also decreasing the peak mechanical stress levels significantly.

With gas temperatures far exceeding the melting point of nickel-base alloys, advanced cooling schemes are essential to meet the desired mission life of turbine airfoils. Naturally, combustion systems produce gas-temperature non-uniformity in the exiting flowfield. Downstream turbine components must be tolerant to the maximum anticipated gas temperatures. On the other hand, excessive use of cooling air reduces engine efficiency and compromises combustor durability. Throughout gas turbine design history it has been the desire of Turbine Aerodynamicists to be able to compute combustor hot streak migration and mixing through multiple turbine airfoil stages. Typically, hot streak migration studies have been performed using (a) mixing-plane models between rotating and stationery domains or (b) unsteady simulations in which the flowpath annulus is represented by a segment containing airfoil counts that are integer multiples in each blade row or (c) Non-Linear Harmonic methods. With the development of highly-parallelized Computational Fluid Dynamic (CFD) codes driving high performance computer clusters simulation of combustor hot streak migration through multiple High Pressure (HP) turbine stages using an unsteady, 360° (full-annulus) model can be achieved. To this end, Honeywell, in collaboration with Numeca Corporation, has performed a study to evaluate the state-of the art for computation of the effect on second-stage HP turbine nozzle metal temperatures of combustor hot streaks migrated through the first-stage of a two-stage HP turbine.

Fans are used in industrial refineries, power generation, petrochemistry, pollution control, etc. These fans can perform in sometimes extreme, mission-critical conditions. The design of fans has historically relied on turbomachinery affinity laws, resulting in oversized machines that are expensive to manufacture and transport. With the increasingly lower CPU cost of fluid modeling, designers can now turn to CFD optimization to produce the necessary machine performance and flow conditions while respecting manufacturing constraints. The objective of this study is to maximize the pressure rise across an industrial fan while respecting manufacturing constraints. First, a 3D scan of the baseline impeller is used to create the CFD model and validated against experimental data. The baseline impeller geometry is then parameterized with 21 free parameters driving the shape of the hub, shroud, blade lean and camber. A fully automated optimization process is conducted using Numeca’s Fine™/Design3D software, allowing for a CPU-efficient Design Of Experiment (DOE) database generation and a surrogate model using the powerful Minamo optimization kernel and data-mining tool. The optimized impeller coupled with a CFD-aided redesigned volute showed an increase in overall pressure rise over the whole performance line, up to 24% at higher mass flow rates compared to the baseline geometry.