Anushka Drescher

Novel approach for tomographic reconstruction of gas concentration distributions in air: Use of smooth basis functions and simulated annealing

Optical remote sensing and iterative computed tomography (CT) can be applied to measure the spatial distribution of gaseous pollutant concentrations. We conducted chamber experiments to test this combination of techniques using an open path Fourier transform infrared spectrometer (OP-FTIR) and a standard algebraic reconstruction technique (ART). Although ART converged to solutions that showed excellent agreement with the measured ray-integral concentrations, the solutions were inconsistent with simultaneously gathered point-sample concentration measurements. A new CT method was developed that combines (1) the superposition of bivariate Gaussians to represent the concentration distribution and (2) a simulated annealing minimization routine to find the parameters of the Gaussian basis functions that result in the best fit to the ray-integral concentration data. This method, named smooth basis function minimization (SBFM), generated reconstructions that agreed well, both qualitatively and quantitatively, with the concentration profiles generated from point sampling. We present an analysis of two sets of experimental data that compares the performance of ART and SBFM. We conclude that SBFM is a superior CT reconstruction method for practical indoor and outdoor air monitoring applications.

Imaging Indoor Tracer-Gas Concentrations with Computed Tomography: Experimental Results with a Remote Sensing FTIR System

This work demonstrates for the first time the feasibility of computed tomography (CT) reconstructions of pollutant concentrations in a real room setting. A remote sensing Fourier transform infrared spectrometer was mounted on a moving base in a controlled ventilation chamber. A passive tracer was released from a point source into the room under constant ventilation conditions. A series of experiments gathered multiple path-averaged measurements in a two-dimensional plane for CT reconstruction. Simultaneous readings were gathered with a multiple-point sampling array for later comparison to the CT reconstructed concentrations. Good qualitative agreement between the reconstruction and point sample data was obtained. Limitations encountered due to the temporal resolution, size, and geometry of the experimental apparatus are clearly surmountable with better instrumentation.

Stationary and time-dependent indoor tracer-gas concentration profiles measured by OP-FTIR remote sensing and SBFM-computed tomography

Measurement of gas concentrations in indoor air using optical remote sensing (ORS) and computed tomography (CT) has been suggested but not thoroughly investigated. We present experiments in which one time-varying and 11 different steady-state tracer-gas concentration profiles were generated in a ventilated chamber and sampled in a horizontal plane by an open-path Fourier transform infrared (OP-FTIR) spectrometer for subsequent CT inversion. CT reconstructions were performed using the recently developed smooth basis function minimization (SBFM) technique. The CT reconstructions were compared with simultaneously gathered point-sample concentration measurements. Agreement between the two sampling methods was qualitatively very good, with concentration profiles generated by both methods showing the same features of peak location and shape. Quantitative agreement was generally good to within 50%. We discuss the sources of discrepancy and suggest directions for future research, especially with regard to monitoring time-dependent processes. With further refinements in the SBFM algorithm and improvements in optical remote sensing hardware, this technique promises to yield rapid and accurate measurements of the spatial distribution of gases in indoor environments.

Computed tomography and optical remote sensing: development for the study of indoor air pollutant transport and dispersion

This thesis investigates the mixing and dispersion of indoor air pollutants under a variety of conditions using standard experimental methods. It also extensively tests and improves a novel technique for measuring contaminant concentrations that has the potential for more rapid, non-intrusive measurements with higher spatial resolution than previously possible. Experiments conducted in a sealed room support the hypothesis that the mixing time of an instantaneously released tracer gas is inversely proportional to the cube root of the mechanical power transferred to the room air. One table-top and several room-scale experiments are performed to test the concept of employing optical remote sensing (ORS) and computed tomography (CT) to measure steady-state gas concentrations in a horizontal plane. Various remote sensing instruments, scanning geometries and reconstruction algorithms are employed. Reconstructed concentration distributions based on existing iterative CT techniques contain a high degree of unrealistic spatial variability and do not agree well with simultaneously gathered point-sample data.

Measurement of tracer gas distributions using an open-path FTIR system coupled with computed tomography

Optical remote sensing and iterative computed tomography (CT) can be combined to measure the spatial distribution of gaseous pollutant concentrations in a plane. We have conducted chamber experiments to test this combination of techniques using an Open Path Fourier Transform Infrared Spectrometer (OP-FTIR) and a standard algebraic reconstruction technique (ART). ART was found to converge to solutions that showed excellent agreement with the ray integral concentrations measured by the FTIR but were inconsistent with simultaneously gathered point sample concentration measurements. A new CT method was developed based on (a) the superposition of bivariate Gaussians to model the concentration distribution and (b) a simulated annealing minimization routine to find the parameters of the Gaussians that resulted in the best fit to the ray integral concentration data. This new method, named smooth basis function minimization (SBFM) generated reconstructions that agreed well, both qualitatively and quantitatively, with the concentration profiles generated from point sampling. We present one set of illustrative experimental data to compare the performance of ART and SBFM.

System for rapid detection and mapping of gas plumes on 100 m scales: examination of some technical and economic issues

We consider the design of a system combining computed tomography and Fourier Transform Infrared Spectroscopy (CT/FTIR) to detect and map the concentration of multicontaminant gas plumes in ambient air over a 100 m square area. Several factors affecting the accuracy of the reconstructed map and the detection limits that can be achieved in the field are discussed. The estimated cost and capabilities of the system are compared with those of a more conventional gas monitoring system that might operate over a similar spatial extent. The paper includes a description of a proposed system that is designed to produce a map of multiple gaseous contaminants with a resolution of 12 m X 12 m in a time of approximately 10 minutes by sequentially measuring the contaminant concentrations along 48 intersecting beam paths and then reconstructing the map using a CT algorithm adapted to detect Gaussian plumes. The optical elements consist of an FTIR mounted on a steerable telescope platform, a second remote steerable mirror platform, and 32 fixed retro-reflectors.

Evaluation of beam-path configurations for use with a monostatic OP-FTIR capable of rapid beam movement for tomographic reconstruction of gas and vapor concentrations in workplaces

Beam path average data from an open path Fourier transform infrared (OP-FTIR) spectrometer can be used to reconstruct 2D concentration maps of the gas and vapor contaminants in workplaces using computed tomographic (CT) techniques. However, a practical limitation arises because many source and detector units are required to produce a sufficient number of intersecting beam paths in order to reconstruct concentration maps. A monostatic OP-FTIR system which is capable of rapid beam movement can be used to eliminate this deficiency. Instead of many source and detector units, a number of the intersecting folded beam paths can be obtained using many flat mirrors and retro-reflectors. We conducted tests of several beam configurations generated for a single scanning FTIR system using 54 flat mirrors and 56 retro-reflectors mounted along the perimeter walls of a typical sized 24 foot by 21 foot test room. The virtual source CT configurations were tested using concentration maps created from tracer gas concentration distributions measured experimentally in a test chamber. Computer simulations of different beam configurations were used to determine the optimal beam geometry. We found that high concentration areas and the general concentration gradient pattern could be resolved from tomographic reconstructions calculated based on 102 folded beam paths. However, the reconstructions showed some effects from noise and peak-smearing artifacts. The noise level could be reduced and the quality of reconstruction maps were improved by using a spline interpolation method to correct for the influence of folded rays. We refer to this approach as a virtual source CT geometry.