Statistical analysis of indoor human exposure by a WLAN source based on polynomial chaos expansions

Zicheng Liu 1 Dominique Lesselier 2 Joe Wiart 1
1 C2M - Chaire Modélisation, Caractérisation et Maîtrise des expositions aux ondes électromagnétiques
2 L2S - Laboratoire des signaux et systèmes
Abstract : Abstract. Human exposure is statistically analyzed in an indoor downlink scenario. Measuring the field emitted by a WLAN source and representing the emitted field by a spherical wave expansion, the whole-body specific absorption rate of a high-resolution human model, named as “Eartha", from the Virtual Classroom is computed with an in-house finite-difference-time-domain code. A surrogate model based on polynomial chaos expansion (PCE) is built for statistical analysis considering the high computational cost of the FDTD code. With the Latin Hypercube sampling method, an experimental design is obtained and the so-called sparse PCE model is then built based on least angle regression. Global sensitivity is analyzed through the computation of Sobol’ indices based on the PCE coefficients. The prediction performance is estimated and results of sensitivity analysis are presented
Keywords : indoor human exposure WLAN source statistical analysis polynomial chaos expansions
Document type :
Conference papers
Complete list of metadatas
https://hal-centralesupelec.archives-ouvertes.fr/hal-01996758
Contributor : Dominique Lesselier <>
Submitted on : Monday, January 28, 2019 - 3:37:54 PM
Last modification on : Thursday, July 4, 2019 - 7:13:40 AM

Identifiers

  • HAL Id : hal-01996758, version 1

Collections

CENTRALESUPELEC | INSMI | SUP_LSS | L2S | TDS-MACS | SUP_SIGNAUX | UNIV-PARIS-SACLAY

Citation

Zicheng Liu, Dominique Lesselier, Joe Wiart. Statistical analysis of indoor human exposure by a WLAN source based on polynomial chaos expansions. 4th Workshop Uncertainty Modeling for Engineering Applications (UMEMA 2018), Dec 2018, Split, Croatia. ⟨hal-01996758⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

55

Contact

support.ccsd.cnrs.fr
hal.support@ccsd.cnrs.fr
Logo CCSD