EM TECHNOLOGY
research2.png
COMPUTATIONAL EM

 

 

EM simulation and visualization of highly complex configurations (> 1e3 CAD parts, > 1e8 unknowns) are performed within just a few minutes (GPU accelerated) using novel robust algorithms

EM simulation and visualization of highly complex configurations: the novel EM-FDTD subgrid engine & GPU computing allows for refining a smart watch on a whole body phantom, allowing a reduction of computation time from one week to four hours.

 

EM simulation and visualization of highly complex configurations (> 1e3 CAD parts, > 1e8 unknowns) are performed within just a few minutes (GPU accelerated) using novel robust algorithms

Efficient modeling of safety/exposure related aspects of Wireless Power Transfer: electrical vehicle charging including two postured anatomical Computable Phantoms.

 

 

 

Computational ElectroMagnetics (CEM)

Background

With the growing trend towards mobility and miniaturization in communications, computing, and medicine, a multitude of interconnected embedded systems and services are increasingly pervading our surrounding environment. Researchers and engineers are constantly challenged to analyze, design, and develop highly efficient devices while satisfying increasing market demands and complying with strict safety guidelines related to the possible health effects of radio-frequency (RF) and low-frequency (LF) electromagnetic (EM) field exposure. The exponential growth of computational power in recent decades has contributed to the expanding role of advanced numerical simulation techniques in efficiently supporting complex R&D processes in research and in industry. In particular, computational electromagnetics (CEM) techniques are vital to the analysis and design of highly complex devices and applications as well as to predict and analyze the interaction mechanisms of electromagnetic fields within complex environments. Today's simulation challenges encompass frequencies ranging from DC to THz, addressing applications such as WPT, In-/On-Body Communication, MR & Implant Safety, EM-Neuron Stimulation, etc. - demanding for dedicated and robust solvers and methodologies.

For more than 15 years, the strong commitment of the IT'IS Foundation to its numerics and simulation R&D efforts has nurtured unique and pacesetting innovations in its core competency of computational electromagnetics.

Selected Past Achievements

  • Developed pioneering algorithms for FDTD local mesh refinement and special sub-cell material treatment (e.g., SIBC) to significantly increase spatial detail at reduced computation times (Chavannes, et al., 2005), (Benkler, et al., 2006), (Schild, et al., 2007) and (Chavannes, ETH Thesis No. 14577, 2002).
  • Conducted thorough analyses of the accuracy of the FDTD algorithm (material interfaces, dispersion) and provided robust correction and improvement methods (Christ, ETH Thesis No. 15057, 2003).
  • Developed and implemented the fastest FDTD solvers currently available using GPU and hybrid GPU/CPU (OpenCL, CUDA) technology (Stefanski, et al., 2011). The IT'IS developed algorithms were integrated into a simulation platform that is the first ever computational tool applying GPU technology.
  • Developed a variety of novel methods for the computational modeling of complex dispersive and non-linear materials (Schild, ETH Thesis No. 17969, 2008).
  • Developed and implemented FDTD rectilinear graded real-time generation of grids and conformal meshing algorithms for robust voxel generation based on 100'000s of complex CAD parts (Benkler et al., 2008).
  • Developed novel methods applying the Generalized Huygens approach to smoothly interconnect very detailed spatial structures embedded within large computational domains (Chavannes, Benkler, et al., 2008).
  • Developed and implemented versatile algorithms for low frequency EM simulations based on a Finite Element (FE) formulation and applied to a graded rectilinear grid. The robust scheme can be applied to complex largely inhomogeneous material distributions, e.g., anatomical human bodies (Bakker, Benkler, Chen et al., 2012).
  • Developed the fastest and most reliable standard compliant methods for the accurate averaging of SAR in complex environments and situations (Crespo, et al., 2010).
  • Developed and implemented various Genetic Algorithm- (GA-) based optimization methods for the fast and robust multi-goal and multi-parameter optimization of highly complex CAD derived structures, e.g., from industry level datasets (Nunez, Chen et al., 2012).
  • Developed the first ever simulation-platform-embedded Alternating Direction Implicit (ADI) FDTD method  to overtune the method inherent time-step and compute complex configurations faster (Benkler, ETH Thesis No. 16969, 2006).
  • Developed solvers and novel toolsets for efficiently simulating complex implants inside the full human body. In addition, developed a comprehensive risk assessment methodology, to determine the specific conditions that would permit an MRI examination for implant-bearing patients (Zastrow, et. al., 2014), (Murbach, ETH Thesis No. 21514, 2013).
  • Developed novel methods to perform physics- (mechanics-) based deformation of models for posing and organ shape modification and applying them to sophisticated high-resolution anatomical models with appropriate meshes (Lloyd, et. al., 2016).
  • Developed novel and highly effective EM-FDTD adaptive subgridding techniques which enable the simulation and optimization of complex In-/On-Body Communication scenarios, featuring micrometer structures embedded within multi-meter environments (Chavannes, et. al., 2016).
  • Developed new solvers and toolsets for the efficient modeling of safety/exposure related aspects of Wireless Power Transfer systems (Chen, et. al., 2014), (Nadakuduti, et. al., 2015).

Next Challenges

  • Keeping up with providing the fastest solvers in the community; developing novel solvers optimized for HPC environments, e.g., using OpenCL/MPI for distributed-memory cluster multi-GPU systems (NVIDIA Pascal based).
  • Extending the solver suite to incorporate other methods, i.e., Finite-Elements (FE) based, operating in the frequency-domain (FD) or on unstructured grids, e.g., to effectively solve general low-frequency (LF) problems.
  • Further improving to bridge the gap between very small structures embedded within large domains, in addition to the local refinement schemes and the Generalized Huygens approach, e.g., by combining adaptive sugridding and conformal FDTD with the latest GPU/HPC technology.
  • Extending the new parameter sweep and optimization engine by incorporating novel methods for DoE and uncertainty analysis.