Deep Learning for Gas Sensing via Infrared Spectroscopy.

Deep learning methods, a powerful form of artificial intelligence, have been applied in a number of spectroscopy and gas sensing applications. However, the speciation of multi-component gas mixtures from infrared (IR) absorption spectra using deep learning remains to be explored. Here, we propose a one-dimensional deep convolutional neural network gas classification model for the identification of small molecules of interest based on IR absorption spectra in flexible user-defined frequency ranges. The molecules considered include ten that are of interest in the atmosphere or in industrial and environmental processes: water vapor, carbon dioxide, ozone, nitrous oxide, carbon monoxide, methane, nitric oxide, sulfur dioxide, nitrogen dioxide, and ammonia. A simulated dataset of IR absorption spectra for mixtures of these molecules diluted in air was generated and used to train a deep learning model. The model was tested against simulated spectra containing noise and was found to provide speciation predictions with accuracy from 82 to 97%. The internal operation of the model was investigated using class activation maps that illustrate how the model prioritizes spectral information for classification. Finally, the model was demonstrated for the prediction of speciation for two synthetic experimental mixture spectra. The proposed model and the dataset generation strategies are generalized and can be implemented for other gases, different frequency ranges, and spectroscopy types. The multi-component speciation method developed herein is the first application of a convolutional neural network model, trained on HITRAN-based simulations, for spectral identification.

TSMC-Net: Deep-Learning Multigas Classification Using THz Absorption Spectra.

The identification of gas mixture speciation from a complex multicomponent absorption spectrum is a problem in gas sensing that can be addressed using machine-learning approaches. Here, we report on a deep convolutional neural network for multigas classification using terahertz (THz) absorption spectra, THz spectra mixture classifier network or TSMC-Net. TSMC-Net has been developed to identify eight volatile organic compounds in mixtures based on their fingerprint rotational absorption spectra in the 220-330 GHz frequency range. A data set consisting of simulated absorption spectra for randomly generated mixtures, with absorption greater than thresholds representing detectable limits and annotated with multiple labels, was prepared for model development. The supervised multilabel classification problem, i.e., the identification of individual gases in a mixture, is converted to a supervised multiclass classification problem via label powerset conversion. The trained model is validated and tested against simulated spectra for gas mixtures, with and without white Gaussian noise. The trained model exhibits high precision, recall, and accuracy for each pure compound. Class activation maps illustrate the complex decision-making process of the model and highlight relevant frequency regions that are needed to identify unique mixtures. Finally, the model was demonstrated against measured THz absorption spectra for pure species and mixtures, acquired using a microelectronics-based THz absorption spectrometer. The data set generation strategy and deep convolutional neural network approach are generalized and can be extrapolated to other spectroscopy types, frequency ranges, and sensors.

Shock tube and chemical kinetic modeling study of the oxidation of 2,5-dimethylfuran.

A detailed kinetic model describing the oxidation of 2,5-dimethylfuran (DMF), a potential second-generation biofuel, is proposed. The kinetic model is based upon quantum chemical calculations for the initial DMF consumption reactions and important reactions of intermediates. The model is validated by comparison to new DMF shock tube ignition delay time measurements (over the temperature range 1300-1831 K and at nominal pressures of 1 and 4 bar) and the DMF pyrolysis speciation measurements of Lifshitz et al. [ J. Phys. Chem. A 1998 , 102 ( 52 ), 10655 - 10670 ]. Globally, modeling predictions are in good agreement with the considered experimental targets. In particular, ignition delay times are predicted well by the new model, with model-experiment deviations of at most a factor of 2, and DMF pyrolysis conversion is predicted well, to within experimental scatter of the Lifshitz et al. data. Additionally, comparisons of measured and model predicted pyrolysis speciation provides validation of theoretically calculated channels for the oxidation of DMF. Sensitivity and reaction flux analyses highlight important reactions as well as the primary reaction pathways responsible for the decomposition of DMF and formation and destruction of key intermediate and product species.

An experimental and kinetic modeling study of the oxidation of the four isomers of butanol.

Butanol, an alcohol which can be produced from biomass sources, has received recent interest as an alternative to gasoline for use in spark ignition engines and as a possible blending compound with fossil diesel or biodiesel. Therefore, the autoignition of the four isomers of butanol (1-butanol, 2-butanol, iso-butanol, and tert-butanol) has been experimentally studied at high temperatures in a shock tube, and a kinetic mechanism for description of their high-temperature oxidation has been developed. Ignition delay times for butanol/oxygen/argon mixtures have been measured behind reflected shock waves at temperatures and pressures ranging from approximately 1200 to 1800 K and 1 to 4 bar. Electronically excited OH emission and pressure measurements were used to determine ignition-delay times. The influence of temperature, pressure, and mixture composition on ignition delay has been characterized. A detailed kinetic mechanism has been developed to describe the oxidation of the butanol isomers and validated by comparison to the shock-tube measurements. Reaction flux and sensitivity analysis illustrates the relative importance of the three competing classes of consumption reactions during the oxidation of the four butanol isomers: dehydration, unimolecular decomposition, and H-atom abstraction. Kinetic modeling indicates that the consumption of 1-butanol and iso-butanol, the most reactive isomers, takes place primarily by H-atom abstraction resulting in the formation of radicals, the decomposition of which yields highly reactive branching agents, H atoms and OH radicals. Conversely, the consumption of tert-butanol and 2-butanol, the least reactive isomers, takes place primarily via dehydration, resulting in the formation of alkenes, which lead to resonance stabilized radicals with very low reactivity. To our knowledge, the ignition-delay measurements and oxidation mechanism presented here for 2-butanol, iso-butanol, and tert-butanol are the first of their kind.

Experimental investigation of toluene + H --> benzyl + H2 at high temperatures.

The reaction of toluene with hydrogen atoms yielding benzyl and molecular hydrogen, C(6)H(5)CH(3) + H --> C(6)H(5)CH(2) + H(2), was investigated using UV laser absorption of benzyl radicals at 266 nm in shock tube experiments. Test gas mixtures of toluene and ethyl iodide, an H-atom source, diluted in argon were heated in reflected shock waves to temperatures ranging from 1256 to 1667 K at total pressures around 1.7 bar. Measurement of laser absorption at 266 nm due to benzyl radicals allowed determination of the rate coefficient of the title reaction, reaction 1. A two-parameter best-fit Arrhenius expression for the rate determinations over the temperature range of these experiments is given by k(1)(T) = 1.33 x 10(15) exp(-14880 [cal/mol]/RT) [cm(3) mol(-1) s(-1)]. With the use of both the high-temperature shock tube measurements reported here and the rate coefficient determination of Ellis et al. (Ellis, C.; Scott, M. S.; Walker, R. W. Combust. Flame 2003, 132, 291) at 773 K the best-fit rate coefficient for reaction 1 can be described using a three-parameter Arrhenius expression by k(1)(T) = 6.47T (3.98) exp(-3384 [cal/mol]/RT) [cm(3) mol(-1) s(-1)].

High-temperature thermal decomposition of benzyl radicals.

The thermal decomposition of the benzyl radical was studied in shock tube experiments using ultraviolet laser absorption at 266 nm for detection of benzyl. Test gas mixtures of 50 and 100 ppm of benzyl iodide dilute in argon were heated in reflected shock waves to temperatures ranging from 1430 to 1730 K at total pressures around 1.5 bar. The temporal behavior of the 266 nm absorption allowed for determination of the benzyl absorption cross-section at 266 nm and the rate coefficient for benzyl decomposition, C6H5CH2 --> C7H6 + H. The rate coefficient for benzyl decomposition at 1.5 bar can be described using a two-parameter Arrhenius expression by k1(T) = 8.20 x 10(14) exp(-40 600 K/T) [s(-1)], and the benzyl absorption cross-section at 266 nm was determined to be sigma(benzyl) = 1.9 x 10(-17) cm2 molecule(-1) with no discernible temperature dependence over the temperature range of the experiments.

Temperature measurement using ultraviolet laser absorption of carbon dioxide behind shock waves.

A diagnostic for microsecond time-resolved temperature measurements behind shock waves, using ultraviolet laser absorption of vibrationally hot carbon dioxide, is demonstrated. Continuous-wave laser radiation at 244 and 266 nm was employed to probe the spectrally smooth CO2 ultraviolet absorption, and an absorbance ratio technique was used to determine temperature. Measurements behind shock waves in both nonreacting and reacting (ignition) systems were made, and comparisons with isentropic and constant-volume calculations are reported.