arXiv subjects

Theofilos Papadopoulos

Publications and source records attributed to Theofilos Papadopoulos.

3 recordsLinked to original sources

arXiv2025

Extension of Simple and Accurate Inductance Estimation for Rectangular Planar Windings

This paper proposes a method to generalize the equations estimating the inductance of square-shape planar windings to rectangle shape. This is done by utilizing the optimal p-norm of the Generalized Mean Value or Power Mean (PM). Three well-established equations with verified accuracy are examined, namely Wheeler, Rosa, and the Monomial, which by definition consider only regular polygons. One critical parameter of the original equations is the outer-side length of the winding, which for the rectangle case, can be substituted by the PM of the two outer-side lengths, without the need for any further modifications. A methodology to select the optimal p-norm for the PM is presented in terms of achieving the best accuracy for this estimation. The selection of the optimal p is based on results from datasets containing more than 2600 simulations of different rectangle-shaped windings. Finally, the estimation accuracy is verified by laboratory measurements for a selection of planar inductors.

eess.SY↗

arXiv2025

Inductance Estimation for High-Power Multilayer Rectangle Planar Windings

This paper proposes a simple and accurate monomial-like equation for estimating the inductance of Multilayer Rectangle-shaped Planar Windings (MLRPWs) for high-frequency, high-power applications. The equation consists of the power product of the geometrical dimensions, raised at individual power coefficients. The coefficients are generated via Multiple Linear Regression (MLR), based on a large set of approximately 6,000 simulated windings, with an 80/20 training/evaluation sample ratio. The resulting mean error value is 0%, with a standard deviation below 1.8%. The accuracy of the inductance estimation is confirmed on several experimental samples, with dimensions both within and outside the initial training dataset.

eess.SY↗

arXiv2024

Adaptive Training of Grid-Dependent Physics-Informed Kolmogorov-Arnold Networks

Physics-Informed Neural Networks (PINNs) have emerged as a robust framework for solving Partial Differential Equations (PDEs) by approximating their solutions via neural networks and imposing physics-based constraints on the loss function. Traditionally, Multilayer Perceptrons (MLPs) have been the neural network of choice, with significant progress made in optimizing their training. Recently, Kolmogorov-Arnold Networks (KANs) were introduced as a viable alternative, with the potential of offering better interpretability and efficiency while requiring fewer parameters. In this paper, we present a fast JAX-based implementation of grid-dependent Physics-Informed Kolmogorov-Arnold Networks (PIKANs) for solving PDEs, achieving up to 84 times faster training times than the original KAN implementation. We propose an adaptive training scheme for PIKANs, introducing an adaptive state transition technique to avoid loss function peaks between grid extensions, and a methodology for designing PIKANs with alternative basis functions. Through comparative experiments, we demonstrate that the adaptive features significantly enhance solution accuracy, decreasing the L^2 error relative to the reference solution by up to 43.02%. For the studied PDEs, our methodology approaches or surpasses the results obtained from architectures that utilize up to 8.5 times more parameters, highlighting the potential of adaptive, grid-dependent PIKANs as a superior alternative in scientific and engineering applications.

cs.LG↗