9 results
Now showing1 - 9 of 9
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Developments in machine learning based modeling and design for electromagnetic compatibility(IEEE, 2025-12-24); ; ; ; Machine learning (ML) has gained traction and emerged as one of most researched topics in recent years. Many engineering disciplines are taking advantage of a multitude of developed ML tools to enhance the engineering workflow, solve complicated problems or speedup design processes. This paper presents an overview on past and present developments of ML-based modeling and design in the field of electromagnetic compatibility (EMC) and the related fields of signal integrity (SI), power integrity (PI), bioelectromagnetics (BioEM) and electromagnetic field (EMF) scans. A critical analysis of recent research on ML applications is carried out to identify the arising opportunities and challenges facing the wide adoption of ML for the modeling and design in EMC. Based on current trends, a vision for ML-based modeling and design in EMC is presented in an attempt to foresee the direction of near-future research. A set of recommendations is formulated to tackle actual challenges facing a wider adoption. ML integration in EMC engineering has a great potential to revolutionize the modeling and design processes, despite the open challenges and the rising complexities of the faced problems.Publicationtype: Journal Article77 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Data-driven path loss estimation in human body communication: enhancing efficiency via parameter prioritization and transfer learningThis work proposes an efficient machine learning approach for path loss estimation in human body communication, focusing on in-body (IB) and on-body (OB) scenarios. Friis' formula and surface wave-based model often have limitations, especially considering variations in electrical properties (EPs), antenna placement, and propagation effects. Here, artificial neural networks (ANNs) are used with parameter prioritization and transfer learning to enhance efficiency. An IB-to-OB link (10 MHz-50 MHz) is analyzed, demonstrating that ANN achieves about 80% accuracy with limited data. OB-to-IB communication at 13.56 MHz, 2.45 GHz, and 5.8 GHz is investigated, incorporating EP and antenna orientation variations. Results show ANN with prioritization reduces input parameters by 70%, maintaining up to 95% accuracy. Additionally, transfer learning enables path loss prediction for 5.8 GHz scenario using a pre-trained 2.45 GHz dataset, achieving over 93% accuracy, significantly reducing computational needs.Publicationtype: Conference Paper35 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Data-driven path loss estimation in human body communication: enhancing efficiency via parameter prioritization and transfer learningThis work proposes an efficient machine learning approach for path loss estimation in human body communication, focusing on in-body (IB) and on-body (OB) scenarios. Friis' formula and surface wave-based model often have limitations, especially considering variations in electrical properties (EPs), antenna placement, and propagation effects. Here, artificial neural networks (ANNs) are used with parameter prioritization and transfer learning to enhance efficiency. An IB-to-OB link (10 MHz-50 MHz) is analyzed, demonstrating that ANN achieves about 80% accuracy with limited data. OB-to-IB communication at 13.56 MHz, 2.45 GHz, and 5.8 GHz is investigated, incorporating EP and antenna orientation variations. Results show ANN with prioritization reduces input parameters by 70%, maintaining up to 95% accuracy. Additionally, transfer learning enables path loss prediction for 5.8 GHz scenario using a pre-trained 2.45 GHz dataset, achieving over 93% accuracy, significantly reducing computational needs.Publicationtype: Conference Paper8 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Full-body vs. head-only modeling: Full wave computational SAR and adaptation of corresponding ANN modelsElectromagnetic compatibility (EMC) analysis is often computationally expensive, with partial modeling and domain-specific approximations commonly employed to improve efficiency, although these simplifications can introduce accuracy trade-offs. To address these challenges, this work focuses on bioelectromagnetic compatibility (Bio-EMC) problems, particularly the Specific Absorption Rate (SAR) calculations, by evaluating SAR in human head tissues using Full-Body and Head-Only models with finite element method (FEM) solvers under plane wave (PW) and near field (NF) exposures at 13.56 MHz. More than 2,000 full wave simulations are conducted, incorporating uncertainties in material properties and exposure angles, with machine learning techniques applied for enhanced analysis. Results show that while model truncation can impact SAR, certain scenarios allow Head-Only data to effectively replace Full-Body data. In these cases, parameter prioritization in artificial neural networks (ANNs) achieves over 90% accuracy while reducing input parameters by up to 70%. For cases where truncation effects are more significant, the ANN trained on Head-Only data is refined using Full-Body data, improving predictive accuracy up to 85% while maintaining computational efficiency. The proposed ANN-based approach enhances both computational efficiency and prediction reliability in Bio-EMC analysis, making it applicable to other emission susceptibility scenarios by reducing system complexity and improving the physical interpretation of results.Publicationtype: Conference Paper13 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Efficient Iterative Data Generation Using Evaluation of Prioritized Input Parameters in ANNs for SAR Prediction in Human Head Models at 13.56 MHzIn this work, an efficient iterative dataset generation strategy is proposed, considering prediction accuracy and high impact input parameters as figures of merit to define a sufficient number of samples for reliable machine learning (ML) results. Parameter prioritization in combination with artificial neural networks (ANNs) is designed for examination of multiparameterized simulation setups in bioelectromagnetic (BEM), aiming to avoid redundant parameters and excessive sample sizes and to provide an alternative to expensive measurements, full-wave simulations, and the limitations of adaptive sampling methods in high-dimensional BEM problems. Specifically, the variation of mass-averaged specific absorption ratio (SAR) in each individual tissue in human head models is studied, considering up to ±\pm 95% uncertainty in a uniform distribution of electrical properties of tissues. Up to 3500 full-wave simulations for seven different scenarios are performed. Utilizing parameter prioritization in ANNs enables high accuracy results with fewer input parameters, allowing improved physical interpretability. By applying this method, the required number of numerical simulations (samples) for an optimal dataset is approximately 5 to 10 times the total number of input parameters. The results of this innovative method demonstrate that the reduced dataset successfully encapsulates the core aspects of the SAR problem under investigation, resulting in ML prediction accuracy surpassing 95% while reducing time and memory consumption by approximately 60%.Publicationtype: Journal Article20 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , SAR prediction for human head models considering dependencies on incident angle of exposure using parameter prioritization in ANNsAccurately characterizing the impact of various parameters involved in bioelectromagnetic (Bio-EM) exposure problems presents computational and experimental difficulties. Measurement methods are often inadequate while computational methods demand costly resources and interpreting results becomes challenging when dealing with numerous parameters. In this study, parameter prioritization in combination with artificial neural networks (ANNs) is employed to predict the specific absorption rate (SAR) in human head tissues with fewer input parameters showcasing improved physics interpretability, computational efficiency, and predictive accuracy. Up to 500 FEM full wave simulations are used to generate data on each human head models. The mass averaged SAR is predicted with over 90% accuracy and using up to 70% fewer input parameters in individual tissues having ±20% electrical properties (EP) uncertainty exposed to a NFC loop antenna with different incident angles. This approach improves understanding of the underlying physics by identifying the most influential parameters.Publicationtype: Conference Paper10 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Physics Inspired Artificial Neural Network Adaptation for SAR Prediction in Bio-EM Problems(2023-09); ; Machine learning (ML) technique is nowadays pop-ular for predicting electromagnetic field exposure, specifically as specific absorption rate (SAR) values, in Bio electromagnetic (Bio-EM) area. Considering the material uncertainty of human tissues, efforts to quantify SAR values mostly rely on 3D full wave simulations, rather than realistic measurements. For precise SAR calculations, a high-resolution human model is often desired and expensive computational resources are required. In this work, an artificial neural network (ANN) as a ML method is employed for SAR prediction in the human head at 13.56 MHz under plane wave illumination. The dependency of SAR values on pm 20% material parameter changes is expected to be linear and separable towards each parameter, inspired by physical inspection of fields. This allows ANN adaptation for effective acceleration of SAR prediction using reduced ANN models. For validation, the proposed ANN modeling uses up to hundreds of full-wave simulations as training data and multiple human head models. This approach achieves a good accuracy of over 95% for SAR prediction. Using the proposed model facilitates fast and accurate predicted results of SAR values for critical situations, which can be compared with standard levels despite the uncertainty in tissue dielectric properties.Publicationtype: Conference Paper21 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , SAR Prediction in Human Head Tissues with Varying Material Parameters Using an Artificial Neural Network Proceedings Article(2023-06); ; Nowadays, machine learning (ML) techniques are becoming popular in bio-electromagnetics. These techniques increase the capability of learning more experience and having more interpretations without explicit programming. In this contribution, artificial neural network (ANN) as a ML method is used to predict the specific absorption rate (SAR) in human head tissues versus parameter variation, during a plane wave field exposure. The dielectric properties of biological tissues are investigated for decades with an uncertainty index in provided values. As reported in literature, biological tissues show variability in dielectric properties up to ± 25% below 100 MHz. Efforts to quantify SAR values, considering tissues properties variation, mostly rely on full-wave simulations with extensive computational resources. An ANN framework is synthesized to facilitate having more results of SAR values in comparison to standard levels, over the tissue dielectric properties uncertainty. The inputs are dielectric properties (s and e) within ±20% variation from nominal values in a specific frequency (13.56 MHz) and local SAR values on each tissue are outputs. The framework is implemented and validated with various datasets of up to 500 full-wave simulation results for four different human head models from spherical to realistic ones. The results indicate up to ±6% and ±12% change on SAR values over ±20% variation of permittivity and conductivity properties respectively. The average accuracy on local SAR prediction by a three layer ANN is about 98% for various human head models within a few minutes implementation and calculation time in comparison to time and memory consuming full wave simulations.Publicationtype: Conference Paper47 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Flexible Numerical Evaluation of Human Head Exposure to a Transmitter Coil for Wireless Power Transfer at 13.56MHzWireless power transfer (WPT) systems are widely applied for brain implants. In this context, the electromagnetic safety of brain tissue exposed to the magnetic and induced electric fields is important. In this contribution, the field exposure levels related to the transmitter (TX) coil of a 13.56 MHz WPT system inside a human head model are investigated. In order to surmount the computation challenges in field detection of WPT structures in human head models, a Huygens' source replacement approach in combination with method of moment (MoM) is utilized. The flexibility of this approach facilitates the simulation in terms of various parameter sweeps such as distance and angle variations. To implement the simulation setup, the TX coil is first stimulated with a 1 V voltage generator in a full-wave simulation and then replaced by a Huygen's source for MoM simulations at 13.56 MHz. Fast and efficient investigations in comparison to other full-wave simulations are achieved. The simulation results show the human head exposure level of the electric field up to 9 dBV/m and the magnetic field up to -36 dBA/m.Publicationtype: Conference Paper31