Feasible adjoint sensitivity technique for EM design optimization

IEEE Trans. Antennas Propagat.

2009

Self Cite

We present a novel algorithm for adjoint sensitivity analysis of time-varying responses that is based on the transmission line modeling (TLM) method. We show that by using at most one extra electromagnetic (EM) simulation, the sensitivities of the complete time domain EM response can be obtained with respect to all designable parameters regardless of their number and regardless of the number of time steps. The complete time domain signature of a certain structure can thus be efficiently and accurately predicted for perturbed structures. Our sensitivities match those obtained using the accurate and time intensive central finite differences (CFD). The adjoint sensitivities are used in the solution of inverse problems through a number of approaches exploiting surrogate model optimization and direct optimization. Our technique enables efficient identification of object position, dimensions and material properties.

Section: Introductionmentioning

confidence: 99%

Adjoint First Order Sensitivities of Transient Responses and Their Applications in the Solution of Inverse Problems

Bakr

Zhao

IEEE Trans. Antennas Propagat.

2009

Self Cite

“…The adjoint-variable method is the most efficient approach to the sensitivity analysis of high-complexity problems modeled numerically [1]- [5]. Its efficiency in terms of computer time is orders of magnitude better compared to derivative estimations using finite-difference or higher-order approximations at the response level.…”

Section: Introductionmentioning

confidence: 99%

“…Its major drawback is the need to compute the system matrix derivatives. The conventional approach of finding analytical expressions of these [3][4], is not feasible for a general-purpose sensitivity solver due to the unrealistic pre-processing effort [5].…”

Section: Introductionmentioning

confidence: 99%

“…Feasible adjoint-based techniques were later proposed where, depending on the nature of the discretization algorithm, approximations of the system-matrix derivatives were attained using either finite differences [5] or discrete step-wise changes [6]- [8]. The former case poses difficulties in implementation since it requires access to the system matrix built by the simulator.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Electromagnetic Sensitivity Analysis of Scattering Parameters Based on the FDFD Method

Zhu

Hasib

2007 International Symposium on Signals, Systems and Electronics

2007

We propose a novel technique to compute response gradients (Jacobians) from frequency-domain field solutions provided by high-frequency electromagnetic (EM) simulations. It is based on our recently developed self-adjoint sensitivityanalysis (SASA) approach where only one EM simulation suffices to obtain both the responses and their gradients in the optimizable-parameter space. Our novel technique exploits the computational efficiency of the SASA while adapting it to the system equations of the frequency-domain finite-difference (FDFD) method. There are three major advantages to this development: (a) the Jacobian computation is completely independent of the simulation engine, its grid and its system equations; (b) the implementation is straightforward and in the form of a post-processing algorithm operating on the exported field solution; (c) it is computationally very efficient-memory and computer-time requirements are negligible compared to those of the simulation itself. The proposed technique drastically reduces the overall time required by field-based optimization processes arising in design and inverse problems as compared to response Jacobians computed via response-level finite differences or parameter sweeps. Its accuracy is verified by comparisons with response-level central finite-difference derivative estimates.Index Terms -Sensitivity analysis, finite-difference frequency-domain method, finite-element method .

“…Recently, approximate sensitivities were considered by Georgieva et al (2002) for frequency-domain solvers on unstructured grids and applications with the MoM. There, the structure is rediscretized at each design iteration for every design parameter to accommodate, at most, 1% to 2% parameter perturbations.…”

Section: Introductionmentioning

confidence: 99%

A Discrete Adjoint Variable Method for Printed-Circuit Board Computer-Aided Design

Ali

INFORMS Journal on Computing

Bakr

2006

W e propose an adjoint-variable method for design sensitivity analysis of printed circuits and antennas where allowable perturbations in the design parameters are of a discrete type. We extend previous work on the sensitivity analysis of waveguide structures, where changes in the design parameters are stepwise, on-grid volumetric perturbations. Here, we explore the feasibility of such an approach in the case of printedcircuit board problems (with open boundaries) where perturbations relate to the shapes elements of infinitesimal thickness. We propose a complex-variable formulation of our approximate sensitivity analysis that improves its computational efficiency. The proposed technique offers significant increases in efficiency, accuracy, and convergence when compared to traditional sensitivity-analysis techniques. Its implementation is straightforward. The response and its gradient with respect to all possible design parameters are computed with at most two full-wave analyses-of the original and the adjoint problems. It operates on a fixed discretization grid where perturbations of grid nodes are not needed. We illustrate our technique through the sensitivity analysis of a microstrip line and a probe-fed printed patch antenna as well as the optimization of a printed Yagi antenna array.