New Technology Radically Speeds the Simulation of Transient Electromagnetic Fields

Transient Electromagnetic SimulationElectric machines, power and electronic transformers and other devices can be better designed and analyzed using transient electromagnetic field simulation. This choice allows engineers to analyze the dynamic system including the non-linear materials, permanent magnets and induced eddy current under a variety of conditions, employing various excitations including the pulsed waveform.

The transient process normally involves solving many time steps in a sequential fashion to calculate saturation, eddy currents, slotting effects and rotor movement in time and space. Because transient electromagnetic analysis requires the computation of many time steps — the process is slow and it limits the size of problem that can be solved — even with the today’s fastest computers. In fact, it is a huge computational undertaking to characterize an electric machine, of any kind, to steady state operation. It can take days or weeks to complete!

ANSYS Maxwell delivers game changing computational speed and capacity for full transient electromagnetic field simulation. The new solving method called Time Decomposition Method (TDM) is used along with specialized solver technology with excellent parallel scalability. The result is an order of magnitude increase in computational speed which leads to a significant increase in the size of simulation that can be solved. This patent pending technology allows engineers to solve all time steps simultaneously instead of sequentially and to distribute the time steps across multiple cores, networked computers and compute clusters.

The Time Decomposition Method is the most effective approach to run large-scale jobs. For example, many time steps required in a transient simulation for an electric motor can be distributed to separate nodes or CPUs and solved simultaneously. Once the individual time steps are solved, their individual solutions can be combined into a full global solution. Maxwell with TDM allows engineers and researchers to solve larger problems more quickly and tackle projects that were previously considered impossible to compute.

The speed and capacity benefits of TDM are scalable across hardware platforms. Even though TDM is mainly targeted for multi-nodes or cluster computing to achieve best performance, it is still applicable to a single multicore computer. Of course, the more hardware available the greater the benefit. The hybrid parallel of TDM method can further leverage high-performance computing (HPC) architectures. Time steps are distributed to multiple nodes and cores by MPI, and extra available cores can be engaged to provide a second layer of parallel computing by shared memory parallel. Benchmark simulations to date have shown excellent scalability in computational speed and capacity as hardware is added.

One example is a simulation performed on 8kW-2000 rpm permanent magnet motor. The magnetic design required 600 time steps with 1.7Mil unknowns (DOFs) on each time step. The simulation without using the Time Decomposition Method lasted 7 days. Applying the TDM method and utilizing the 396 available cores reduced the simulation from 7 days to 2 hours and 39 minutes!

These computational speed and capacity breakthroughs make transient electromagnetic field simulation a viable design tool instead of a final verification tool. Simulations that previously required weeks of simulation time are easily solved in a matter of hours. Critical transient behavior can be assessed during early design stages allowing engineers to acquire design information early in the design stage reducing the risk of project delays and late stage design changes.

For more information and detailed benchmark studies visit our site. There’s also a whitepaper, entitled ANSYS Maxwell and SGI® UV™ 3000 Maximizes Electromagnetic Computation Throughput, that you may be interested in reading that has a companion on-demand webinar.

One thought on “New Technology Radically Speeds the Simulation of Transient Electromagnetic Fields

  1. Nice! I’m also approaching the same problem from a slightly different angle (HB-FEM) for my doctoral studies, so I’d be interested reading a bit more about this TDM.

    Have you published anything yet, or do you still have some patent stuff to sort out?

Comments are closed.