An analysis of prior information in Bayesian tomographic reconstruction

Samuel J. Grauer, Paul J. Hadwin, Kyle J. Daun

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations


Line-of-sight-attenuation chemical species tomography is a diagnostic in which the spatial distribution of a gaseous species is reconstructed from line-of-sight light-absorption measurements. In this approach, the measurement field is discretized into pixels wherein the species concentration of interest is presumed constant. Because the number of pixels needed to resolve the spatial features of interest almost always exceeds the number of measurement paths, additional assumptions about the distribution must be incorporated into the analysis to identify a unique solution. This paper presents a Bayesian approach to tomographic reconstruction, which formalizes the distinct roles of measurement data and prior information in the construction of an a posteriori distribution estimate. This technique was tested on a large-eddy simulation of a turbulent free-shear methane jet. Three forms of prior information were tested, ordered from most-to-least informative: the spatial covariance data from the simulation; a squared-exponential approximation of the spatial covariance; and a first-order Tikhonov matrix, which operated as a basic spatial-smoothness prior. Preliminary results show reconstruction accuracy improves with increasingly informative priors.

Original languageEnglish (US)
Title of host publicationProceedings of the 1st Thermal and Fluid Engineering Summer Conference, TFESC 2015
PublisherBegell House Inc.
Number of pages11
ISBN (Electronic)9781567004311
StatePublished - 2015
Event1st Thermal and Fluid Engineering Summer Conference, TFESC 2015 - New York City, United States
Duration: Aug 9 2015Aug 12 2015

Publication series

NameProceedings of the Thermal and Fluids Engineering Summer Conference
ISSN (Electronic)2379-1748


Conference1st Thermal and Fluid Engineering Summer Conference, TFESC 2015
Country/TerritoryUnited States
CityNew York City

All Science Journal Classification (ASJC) codes

  • Mechanical Engineering
  • Electrical and Electronic Engineering
  • Fluid Flow and Transfer Processes
  • Condensed Matter Physics
  • Energy Engineering and Power Technology
  • Renewable Energy, Sustainability and the Environment


Dive into the research topics of 'An analysis of prior information in Bayesian tomographic reconstruction'. Together they form a unique fingerprint.

Cite this