Computational electromagnetics and electromagnetic inverse imaging

点击翻译按钮获取中文摘要

This five year NSERC Discovery Grant proposal is on high-performance computational electromagnetics (CEM) and EM inverse imaging (EMII) algorithms for biomedical applications (e.g., breast cancer imaging) and the structural-health monitoring of civil infrastructure (e.g., detection of rebar corrosion in bridge decks). Through prior research, we now have a finite-volume time-domain (FVTD) code that solves EM problems on arbitrary flat-faceted volumetric unstructured meshes which executes with a user-specifiable arbitrary order of accuracy. A new integral equation mesh truncation scheme allows us to terminate the mesh very close to the geometrical structures being modeled. This FVTD algorithm, as well as other algorithms we'll investigate as part of this proposal, will be used to (1) synthesize complex EM scattering data for our EMII research, and (2) as the forward-scattering part of the new EMII algorithms we will develop. This will require that we parallelize our CEM algorithms and implement them on the grid-computers that we currently have available. We've already investigated several EMII algorithms on scattering data collected with a vector network analyzer based ground penetrating radar system we've constructed. This has required fundamental investigations into the EM inverse problem and we've recently published a new nonlinear regularization technique based on both the distorted Born iterative technique. For the proposed research, high-resolution techniques, recently developed in the inverse-methods community, will be investigated. These include (1) the Contrast Source Inversion method, and (2) regularization methods based on the L-curve and the Normalized Cumulative Periodogram of the residual. These will be applied to the data collected using our existing GPR system as well as a new system we are developing for imaging human phantoms. Both systems will be expanded to collect transient-pulse scattering data and we will design the required transient-pulse antennas. We will also research ways of implementing the EMII techniques directly in the time-domain.