Tim O’Connell with adviser P. T. Krein
The transportation industry is currently making a large push to “electrify” its fleets by developing hybrid electric vehicles (HEVs) and “more-electric” aircraft. The U.S. military is also funding several projects aimed at producing an all-electric battleship. Such endeavors, and many others across industry, require improved, specialized electric drive systems which utilize advanced electric machines. For these, “off-the-shelf” machines distinguished purely by their steady-state ratings cannot meet the exacting standards required without sacrificing performance. Thus, the time is ripe for engineers to improve the design tools needed to facilitate this industry-wide transformation.
Our research in the past year has focused on identifying bottlenecks in the existing electric drive system design process and then examining and categorizing alternative algorithms to reduce these bottlenecks. We have looked more closely and quantitatively at the computing time needed to analyze electric machines using commercial large-scale numerical field analysis software, and have applied Moore’s law to this data to forecast the time-horizon for design-by-iterative-analysis. Moore’s law predicts that if historical trends hold into the foreseeable future, computer processing speed will continue to double approximately every two years. Figure 1 shows how long Moore’s law predicts that we must wait until a large-scale numerical field analysis tool like 3D finite element analysis (FEA) is able to simulate a machine in real time (i.e., 1 s of simulation time equals 1 s of processor time). For varying system excitation frequencies, this wait time is plotted against the real time needed today to conduct one machine field analysis. As shown, even if an analysis tool can solve a 3D field in one minute today, Moore’s Law predicts that we still must wait at least twenty-five years before a time-domain simulation of even a 60 Hz system is feasible in real time.
We have identified several possible routes to shrink this wait time, including improving existing algorithms, exploring alternative algorithms that are well established in other engineering disciplines but not traditionally applied to complex electric drive system simulation, intelligently combining multiple algorithms into “hybrid” methods that exploit the unique strengths of each, and using alternative hardware implementations such as parallel computing on clusters or multi-core processors. Our current research focus is exploring and categorizing alternative algorithms to synthesize useful hybrid methods. One result of this work can be seen in the proposed hybrid design strategy diagrammed in Figure 2. Starting with a broadly defined 3D electric machine design space, first the Schwarz-Christoffel (S-C) method is used to rapidly reduce it to a more manageable size. The S-C method has been found to quickly evaluate many designs without sacrificing information about the rotor and stator slot geometries. Next, magnetic equivalent circuit (MEC) models and the boundary element method (BEM), aided by a fast multipole method (FMM) accelerator algorithm, are used in a design-by-iterative-analysis loop to further narrow the design space to a few worthy candidates. The final designs are analyzed with the numerically intensive full 3D FEA for verification. Other algorithms not yet identified or examined can easily be incorporated into this hybrid design loop as they become available.
This work is funded by the Grainger Center for Electric Machinery and Electromechanics.