Introduction

Development of modern microwave devices largely exploits full-wave electromagnetic (EM) simulations. Yet, simulation-driven design may be problematic due to the incurred CPU expenses. Addressing the high-cost issues stimulated the development of surrogate modeling methods. Among them, data-driven techniques seem to be the most widespread owing to their flexibility and accessibility. Nonetheless, applicability of approximation-based modeling for real-world microwave components is hindered by a high nonlinearity of the system characteristics, dimensionality issues, and broad ranges of operating parameters the model should cover to make it practically useful. Performance-driven modeling frameworks deliver a partial mitigation of these problems through appropriate spatial orientation of the metamodel domain, which only encapsulates high-quality designs and not the entire space. Unfortunately, the initial model setup cost is high, as defining the domain requires database designs that need to be a priori acquired. This paper introduces a novel approach, where the database designs are replaced by random observables, and dimensionality of the domain is reduced using spectral analysis thereof. The major contributions of the work include implementation of the explicit dimensionality reduction of the confined surrogate model domain and introducing this concept into a complete cost-efficient framework for modeling of microwave components. Comprehensive benchmarking demonstrates excellent performance of the introduced framework, both in terms of predictive power of the rendered surrogates, their scalability properties, as well as low computational overhead associated with the model setup.

Section: Reference-design-free Domain-confined Modelingmentioning

confidence: 99%

Section: Reference-design-free Domain-confined Modelingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Reduced-cost microwave modeling using constrained domains and dimensionality reduction

Koziel,

Pietrenko-Dabrowska,

Ullah

2023

“…In practice, utilizing general-purpose modeling techniques poses problems for devices featuring more than a few geometry parameters 70 – 72 . The mitigation methods include domain confinement 73 , 74 , incorporation of variable-resolution EM simulations 75 , as well as the response feature methodology 76 . The latter benefits from a weakly-nonlinear dependence of the coordinates of appositely singled out characteristic points of antenna responses on the geometry parameters (as opposed to the complete responses), which allows—upon reformulation of the design problem with the use of response features—for a faster convergence of the optimization process 77 , or a reduction of the number of training samples (in the context of surrogate modeling 78 ).…”

Section: Introductionmentioning

confidence: 99%

On nature-inspired design optimization of antenna structures using variable-resolution EM models

Kozieł

Pietrenko‐Dabrowska

2023

Numerical optimization has been ubiquitous in antenna design for over a decade or so. It is indispensable in handling of multiple geometry/material parameters, performance goals, and constraints. It is also challenging as it incurs significant CPU expenses, especially when the underlying computational model involves full-wave electromagnetic (EM) analysis. In most practical cases, the latter is imperative to ensure evaluation reliability. The numerical challenges are even more pronounced when global search is required, which is most often carried out using nature-inspired algorithms. Population-based procedures are known for their ability to escape from local optima, yet their computational efficiency is poor, which makes them impractical when applied directly to EM models. A common workaround is the utilization of surrogate modeling techniques, typically in the form of iterative prediction-correction schemes, where the accumulated EM simulation data is used to identify the promising regions of the parameter space and to refine the surrogate model predictive power at the same time. Notwithstanding, implementation of surrogate-assisted procedures is often intricate, whereas their efficacy may be hampered by the dimensionality issues and considerable nonlinearity of antenna characteristics. This work investigates the benefits of incorporating variable-resolution EM simulation models into nature-inspired algorithms for optimization of antenna structures, where the model resolution pertains to the level of discretization density of an antenna structure in the full-wave simulation model. The considered framework utilizes EM simulation models which share the same physical background and are selected from a continuous spectrum of allowable resolutions. The early stages of the search process are carried out with the use of the lowest fidelity model, which is subsequently automatically increased to finally reach the high-fidelity antenna representation (i.e., considered as sufficiently accurate for design purposes). Numerical validation is executed using several antenna structures of distinct types of characteristics, and a particle swarm optimizer as the optimization engine. The results demonstrate that appropriate resolution adjustment profiles permit considerable computational savings (reaching up to eighty percent in comparison to high-fidelity-based optimization) without noticeable degradation of the search process reliability. The most appealing features of the presented approach—apart from its computational efficiency—are straightforward implementation and versatility.

“…Acceleration methods include the incorporation of techniques for fast evaluation of circuit response gradients (adjoint sensitivities 30 , 31 , mesh deformation 32 , parallelization 33 ), replacing numerical derivatives by updating formulas 34 , as well as utilization of sparse sensitivity updates 35 – 37 . In a more generic setting, surrogate-assisted procedures have been gaining considerable attention 38 – 41 , both in the context of physics-based (space mapping 42 , adaptive response scaling 43 , manifold mapping 44 ), and data-driven models (radial basis functions 45 , kriging 46 , artificial neural networks 47 , ensemble learning 48 , support vector regression 49 ), along with variable-resolution methods (co-kriging 50 , Bayesian model fusion 51 ). Surrogate-based methods are applied for both local 52 , 90 – 93 and global optimization 53 – 55 , but also multi-criterial design 56 – 59 , and uncertainty quantification 60 – 62 .…”

Section: Introductionmentioning

confidence: 99%

Rapid and reliable re-design of miniaturized microwave passives by means of concurrent parameter scaling and intermittent local tuning

Kozieł

Pietrenko‐Dabrowska

2023

Re-design of microwave passive components for the assumed operating frequencies or substrate parameters is an important yet a tedious process. It requires simultaneous tuning of relevant circuit variables, often over broad ranges thereof, to ensure satisfactory performance of the system. If the operating conditions at the available design are distant from the intended ones, local optimization is typically insufficient, whereas global search entails excessive computational expenses. The problem is aggravated for miniaturized components, typically featuring large numbers of geometry parameters. Furthermore, owing to their tightly-arranged layouts, compact structures exhibit considerable cross-coupling effects. In order to reliably evaluate electrical characteristics under such conditions full-wave electromagnetic (EM) analysis is mandatory. Needless to say, EM-driven design over broad ranges of operating frequencies is an arduous and costly endeavor. In this paper, we introduce a novel procedure for rapid and reliable re-design of microwave passives. Our methodology involves concurrent scaling of geometry parameters interleaved with local (gradient-based) tuning. The scaling stage allows for low-cost relocation of the operating frequencies of the circuit, whereas the optimization stage ensures continuous (iteration-wise) alignment of the performance figures with their target values. The presented framework is validated using several miniaturized microstrip couplers, re-designed over extended ranges of the center frequencies. For all considered structures, satisfactory designs are successfully identified despite the initial designs being distant from the targets, whereas local tuning turns out to be demonstrably inferior. Apart from its efficacy, one of the most important advantages of the proposed framework is its simplicity, and the lack of problem-dependent control parameters.