University Grant on Eddy Currents May Impact Machine Design
University researchers are developing a new, more powerful method for simulations of eddy currents that would have a major impact on the design of electrical machines.
Funded by a National Science Foundation-supported grant, “Pfi-tt: A Parallel Computing Engine for Simulation of Complex Multi-Scale Systems,” started August 2018 and concludes at the end of July 2021. The grant is worth $251,164 and looks to develop novel methods to simulate complex systems that are difficult or unpractical to simulate with existing finite element (FEM) tools.
The capabilities of state of the art commercially available FEM tools have become a bottleneck for engineers and scientists working at the boundaries of what the tools can solve. Several important problems cannot be properly simulated with existing software or would require prohibitively long simulation time to reach a single solution.
The proposed research involves a novel approach for the transient simulation of multi-scale systems using Monte Carlo methods. The approach enables parallelization in GPU platforms, leading to a potential breakthrough reduction in simulation time, compared to standard FEM simulation. As a practical example, the simulation of eddy currents in normal conducting coils is being addressed. Hector Gutierrez, mechanical and civil engineer professor, is the primary investigator.
In the presence of an alternating current (AC) magnetic field, parasitic electric currents, known as eddy currents, are induced in any conducting material exposed to the AC field. Accurate modeling of eddy currents has significant applications in engineering and science, such as measuring the magnetic field generated by the induced eddy current to use as a non-destructive technique to detect defects in a metal part. Eddy currents are also used in a non-standard braking system. Unlike mechanical brakes, which are based on friction and kinetic energy, eddy current brakes rely on electromagnetism to stop objects from moving. The formation of eddy currents in the copper substrate of superconducting magnets is also an essential factor in the design of superconducting coils.
Accurate and timely simulation of eddy currents has potential broad impacts in engineering design, in electrical machines. Eddy current research may also have a beneficial impact on the environment, Gutierrez noted.
“There is a significant trend to replace fossil fuels with electric power in propulsion of cars, ships and aircraft,” Gutierrez said. “The study and simulation of eddy currents is more and more important in this context.”
Underwater or underground detection of metals is also a promising commercial application of eddy current-based sensors. To detect the presence of metals, AC magnetic fields can be used to induce eddy currents in electrically conductive materials. The induced eddy currents produce their own magnetic field, which can be detected to determine the presence of conductive materials. The induced eddy currents depend on several physical parameters of the detected sample, such as conductivity, size and orientation.
Most analytical and numerical approaches to simulate eddy currents are mainly based on finite element methods (FEM) or hybrid methods that employ FEM. Although FEM is a well-established technique for simulation of boundary value problems, it also has several well-known limitations. Gutierrez and his team have developed a new method called Effective Floating Volume (EFV) to solve multiphysics problems related to heat transfer, joule heating and electric current sharing.
EFV is a mesh-free method that not only facilitates the definition of the geometry using random points but also does not require matrix inversion or solution of a system of equations.
Adequate meshing of complex geometries is a critical task in FEM as it directly impacts the accuracy of the results. Complex components with multiscale geometries require significant and possibly prohibitive effort to define the domain by mesh elements. Intricate parts often require very fine meshing of the domain to achieve acceptable accuracy and increasing the number of mesh elements results in very large stiffness and mass matrices and subsequently longer simulation times and significantly larger memory resources.
Monte Carlo methods remain a powerful alternative for the simulation of complex multi-scale systems that cannot be properly addressed by existing finite element tools.