Search NASA⌕ Search

Engineering topics

Muralidhar, Ramachandran

Publications and source records attributed to Muralidhar, Ramachandran.

Heuristic based analytics for gas leak source identification

Heuristic-based techniques for gas leak source identification are provided. In one aspect, a method for identifying a location of a gas leak source includes: obtaining gas sensor data and wind data synchronously from a gas leak detection system having a network of interconnected motes comprising gas sensors and wind sensors, with the gas sensors arranged around possible gas leak sources in a given area of interest; identifying the location of the gas leak source using the gas sensor data and wind data; and determining a magnitude of gas leak from the gas leak source using the location of the gas leak source and a distance d between the location of the gas leak source and a select one of the gas sensors from which the gas sensor data was obtained. A gas leak detection system is also provided.

03 NATURAL GAS↗

Field Deployment of a Portable Optical Spectrometer for Methane Fugitive Emissions Monitoring on Oil and Gas Well Pads

We present field deployment results of a portable optical absorption spectrometer for localization and quantification of fugitive methane (CH 4 ) emissions. Our near-infrared sensor targets the 2$ν_3$ R(4) CH 4 transition at 6057.1 cm −1 (1651 nm) via line-scanned tunable diode-laser absorption spectroscopy (TDLAS), with Allan deviation analysis yielding a normalized 2.0 ppmv∙Hz −1/2 sensitivity (4.5 × 10 −6 Hz −1/2 noise-equivalent absorption) over 5 cm open-path length. Controlled CH 4 leak experiments are performed at the METEC CSU engineering facility, where concurrent deployment of our TDLAS and a customized volatile organic compound (VOC) sensor demonstrates good linear correlation (R 2 = 0.74) over high-flow (>60 SCFH) CH 4 releases spanning 4.4 h. In conjunction with simultaneous wind velocity measurements, the leak angle-of-arrival (AOA) is ascertained via correlation of CH 4 concentration and wind angle, demonstrating the efficacy of single-sensor line-of-sight (LOS) determination of leak sources. Finally, source magnitude estimation based on a Gaussian plume model is demonstrated, with good correspondence (R 2 = 0.74) between calculated and measured release rates.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Localization and Quantification of Trace-gas Fugitive Emissions Using a Portable Optical Spectrometer

We present a portable optical spectrometer for fugitive emissions monitoring of methane (CH4). The sensor operation is based on tunable diode laser absorption spectroscopy (TDLAS), using a 5 cm open path design, and targets the 2ν3 R(4) CH4 transition at 6057.1 cm-1 (1651 nm) to avoid cross-talk with common interfering atmospheric constituents. Sensitivity analysis indicates a normalized precision of 2.0 ppmv∙Hz-1/2, corresponding to a noise-equivalent absorption (NEA) of 4.4×10-6 Hz-1/2 and minimum detectible absorption (MDA) coefficient of αmin = 8.8×10-7 cm-1∙Hz-1/2. Our TDLAS sensor is deployed at the Methane Emissions Technology Evaluation Center (METEC) at Colorado State University (CSU) for initial demonstration of single-sensor based source localization and quantification of CH4 fugitive emissions. The TDLAS sensor is concurrently deployed with a customized chemi-resistive metal-oxide (MOX) sensor for accuracy benchmarking, demonstrating good visual correlation of the concentration time-series. Initial angle-of-arrival (AOA) results will be shown, and development towards source magnitude estimation will be described.

03 NATURAL GAS↗

Multi-model blending

A method and a system to perform multi-model blending are described. The method includes obtaining one or more sets of predictions of historical conditions, the historical conditions corresponding with a time T that is historical in reference to current time, and the one or more sets of predictions of the historical conditions being output by one or more models. The method also includes obtaining actual historical conditions, the actual historical conditions being measured conditions at the time T, assembling a training data set including designating the two or more set of predictions of historical conditions as predictor variables and the actual historical conditions as response variables, and training a machine learning algorithm based on the training data set. The method further includes obtaining a blended model based on the machine learning algorithm.

97 MATHEMATICS AND COMPUTING↗