pith. machine review for the scientific record. sign in

arxiv: 2605.01306 · v1 · submitted 2026-05-02 · ⚛️ physics.optics · cs.LG· physics.app-ph

Recognition: unknown

Machine Learning Enhanced Laser Spectroscopy for Multi-Species Gas Detection in Complex and Harsh Environments

Mohamed Sy

Pith reviewed 2026-05-09 18:30 UTC · model grok-4.3

classification ⚛️ physics.optics cs.LGphysics.app-ph
keywords laser absorption spectroscopymachine learningmulti-species gas detectioncombustion diagnosticsautoencodersblind source separationinterference mitigationvolatile organic compounds
0
0 comments X

The pith

Machine learning models integrated with laser absorption spectroscopy enable real-time detection of multiple gas species in interfering and dynamic environments even when reference data is incomplete.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops and tests several machine learning approaches to overcome limits in conventional laser absorption spectroscopy when measuring gas mixtures under high-speed, noisy, or overlapping conditions. Deep denoising autoencoders clean signals from shock-tube pyrolysis tests, while frameworks like HT-SIMNet and UnblindMix separate species contributions without needing complete calibration libraries or reference spectra for every component. Additional techniques recover weak signals from minor species and provide certifiable classification for volatile organics. A reader should care because these methods could support reliable monitoring in combustion, environmental, and industrial settings where traditional spectroscopy fails due to unknown interferents or changing conditions.

Core claim

The central claim is that deep denoising autoencoders improve trace-species detection limits in high-speed hydrocarbon pyrolysis, HT-SIMNet isolates species via spectral augmentation and Noise2Noise-style training without full references, UnblindMix reconstructs concentrations and signatures from mixtures of up to eight components, first-derivative convolution features highlight masked weak absorbers, and randomized-smoothing Voigt perturbation yields certifiable VOC classification. All are experimentally validated on shock-tube and mixture data, establishing a route to interference-resilient, reference-free multi-species gas sensing.

What carries the argument

ML models including denoising autoencoders, structured unsupervised networks, blind source separation autoencoders, derivative feature engineering, and certifiable smoothing classifiers applied directly to laser absorption spectra to handle noise, unknown interferents, and missing references.

If this is right

  • Higher-fidelity signals allow lower detection limits for trace hydrocarbons during rapid pyrolysis.
  • Unknown species interference can be mitigated in reactive flows without exhaustive calibration libraries.
  • Mixture concentrations and pure spectra can be recovered directly from observed data for up to eight components.
  • Weakly absorbing species can be recovered when masked by stronger overlapping absorbers.
  • Classification of volatile organic compounds can be made certifiably robust to condition variations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Hardware integration of these models could enable closed-loop control in industrial combustion systems.
  • The blind separation approach might apply to other spectroscopic domains such as atmospheric remote sensing where species lists are incomplete.
  • Combining the methods with different laser sources or field-deployed sensors would test whether the reference-free property holds beyond laboratory shock tubes.

Load-bearing premise

The trained machine learning models will continue to work accurately when applied to new harsh environments, unknown gas species, or different hardware configurations not represented in the shock-tube and mixture validation tests.

What would settle it

Apply one of the trained models to a new shock-tube experiment or mixture containing previously unseen species or interferents and measure whether detection accuracy or separation quality drops below conventional spectroscopic baselines.

Figures

Figures reproduced from arXiv: 2605.01306 by Mohamed Sy.

Figure 1.1
Figure 1.1. Figure 1.1: Major application areas for gas sensing technologies. view at source ↗
Figure 1.2
Figure 1.2. Figure 1.2: Conventional techniques for multi-species gas sensing and their limitations. view at source ↗
Figure 1.3
Figure 1.3. Figure 1.3: Gas-phase infrared spectra of target species obtained from the PNNL database view at source ↗
Figure 1.4
Figure 1.4. Figure 1.4: Typical machine learning framework for gas sensing applications. view at source ↗
Figure 2.1
Figure 2.1. Figure 2.1: Schematic illustration of the Beer–Lambert law in gas sensing. view at source ↗
Figure 3.1
Figure 3.1. Figure 3.1: Optical schematic of the sensor through the shock tube cross-section. Red view at source ↗
Figure 3.2
Figure 3.2. Figure 3.2: (a) Spectra of methane, ethane, ethylene, acetylene, propane, propene, and view at source ↗
Figure 3.3
Figure 3.3. Figure 3.3: A representative shock experiment of CH4/Ar. The red region indicates pre view at source ↗
Figure 3.4
Figure 3.4. Figure 3.4: Measured temperature-dependent absorption cross-sections of methane, ethane, view at source ↗
Figure 3.5
Figure 3.5. Figure 3.5: (a): Measured absorbance signals during the pyrolysis of 2% ethane/Ar at T = view at source ↗
Figure 3.6
Figure 3.6. Figure 3.6: Flow chart of the overall process followed in this work including the training view at source ↗
Figure 3.7
Figure 3.7. Figure 3.7: Results of signal cleaning with DDAEs. Blue lines show simulated noisy view at source ↗
Figure 3.8
Figure 3.8. Figure 3.8: Noisy composite absorbance spectra during the pyrolysis of 2% ethane/Ar at view at source ↗
Figure 3.9
Figure 3.9. Figure 3.9: Mole fraction time-histories during the pyrolysis of 2% ethane/Ar at (a) 1200 view at source ↗
Figure 3.10
Figure 3.10. Figure 3.10: Mole fraction time-histories during the pyrolysis of 2% propane in argon at view at source ↗
Figure 4.1
Figure 4.1. Figure 4.1: Schematic of the experimental setup, including the ICL laser source, heated view at source ↗
Figure 4.2
Figure 4.2. Figure 4.2: Absorbance spectra of the target species under conditions (T = 295 K, P = 1 view at source ↗
Figure 4.3
Figure 4.3. Figure 4.3: Absorbance spectra of a clean mixture (black) alongside a pair of corrupted view at source ↗
Figure 4.4
Figure 4.4. Figure 4.4: Schematic of the end-to-end training and testing workflow for HT-SIMNet and view at source ↗
Figure 4.5
Figure 4.5. Figure 4.5: Comparison of corrupted spectra vs predicted spectra by HT-SIMNet (dashed view at source ↗
Figure 4.6
Figure 4.6. Figure 4.6: Comparison of predicted mole fractions for methane, ethane, ethylene, and view at source ↗
Figure 4.7
Figure 4.7. Figure 4.7: Stacked bar plot comparing the mean absolute error of different augmentation view at source ↗
Figure 4.8
Figure 4.8. Figure 4.8: Absorbance spectra collected during n-butane pyrolysis at different times, view at source ↗
Figure 4.9
Figure 4.9. Figure 4.9: Predicted concentrations of methane and ethylene during n-butane pyrolysis. view at source ↗
Figure 5.1
Figure 5.1. Figure 5.1: Experimental setup view at source ↗
Figure 5.2
Figure 5.2. Figure 5.2: Measured absorption spectra of the target species at T = 295 K, P = 1 atm, L = 21 cm, and mole fractions of 1.1% (methane), 0.5% (ethylene), 0.28% (ethane), 1% (propyne), 0.8% (n-butane), and 0.8% (iso-butane) view at source ↗
Figure 5.3
Figure 5.3. Figure 5.3: Flowchart of the training process for the UnblindMix model view at source ↗
Figure 5.4
Figure 5.4. Figure 5.4: Comparison between a clean mixture spectrum and an augmented spectrum after applying synthetic training data transformations view at source ↗
Figure 5.5
Figure 5.5. Figure 5.5: Measured temperature-dependent absorption cross-sections of methane, ethane, ethylene and propyne over T = 298 – 923 K and P = 1 atm view at source ↗
Figure 5.6
Figure 5.6. Figure 5.6: Comparison of reconstructed spectra and residuals for NMF and UnblindMix view at source ↗
Figure 5.7
Figure 5.7. Figure 5.7: Comparison of reconstructed spectra and residuals for NMF and UnblindMix view at source ↗
Figure 5.8
Figure 5.8. Figure 5.8: Comparison of measured spectra and those predicted by UnblindMix. Mea view at source ↗
Figure 5.9
Figure 5.9. Figure 5.9: Comparison of measured mole fractions and those predicted by AE-BSS (Un view at source ↗
Figure 5.10
Figure 5.10. Figure 5.10: Effect of the initial guess on model convergence. view at source ↗
Figure 5.11
Figure 5.11. Figure 5.11: Effect of training data size on model convergence. Validation loss is plotted view at source ↗
Figure 5.12
Figure 5.12. Figure 5.12: Measured reference spectra vs. predicted by UnblindMix at 923 K during view at source ↗
Figure 5.13
Figure 5.13. Figure 5.13: (a) Actual vs. predicted mole fraction time histories by UnblindMix during view at source ↗
Figure 5.14
Figure 5.14. Figure 5.14: Measured reference spectra vs. predicted by UnblindMix at 923 K during py view at source ↗
Figure 5.15
Figure 5.15. Figure 5.15: (a) Actual vs. predicted mole fraction time histories by UnblindMix during view at source ↗
Figure 6.1
Figure 6.1. Figure 6.1: Measured absorption spectra of the target species at T = 295 K, P = 1 atm, L = view at source ↗
Figure 6.2
Figure 6.2. Figure 6.2: Schematic of the optical sensor setup. 6.3 Spectral feature engineering 6.3.1 Feature transformation process Feature transformation is a common technique in machine learning designed to highlight new features and improve model learning efficiency[162]. In the context of spectral feature engineering, this can be useful particularly for addressing the challenge of detecting species with low absorbance that… view at source ↗
Figure 6.3
Figure 6.3. Figure 6.3: ((a) Original measured spectra of C1-C3 hydrocarbons, where ethylene ab￾sorbance is kept very low (0.05), including a mixture with ethane and propyne at 1% and ethylene at 200 ppm. (b) Feature engineered spectra highlighting key features after spectral transformation. 6.3.2 Model description Convolutional neural networks (CNNs) have proven effective in extracting valuable infor￾mation from corrupted spec… view at source ↗
Figure 6.4
Figure 6.4. Figure 6.4: Flowchart of the standard and feature-engineered models. view at source ↗
Figure 6.5
Figure 6.5. Figure 6.5: Flowchart of the training process: the top light yellow box represents the view at source ↗
Figure 6.6
Figure 6.6. Figure 6.6: MSE loss of the standard and feature engineered models. view at source ↗
Figure 6.7
Figure 6.7. Figure 6.7: Allan deviation plot of the sensor. Optimal integration time is 8.32 seconds view at source ↗
Figure 6.8
Figure 6.8. Figure 6.8: Comparison of predicted mole fractions with manometric (actual) values for view at source ↗
Figure 6.9
Figure 6.9. Figure 6.9: Effect of noise on species prediction view at source ↗
Figure 7.1
Figure 7.1. Figure 7.1: Normalized spectra of twelve VOCs at P = 0.5 Torr and P = 16 Torr view at source ↗
Figure 7.2
Figure 7.2. Figure 7.2: Effect of augmentations across varied FWHMs. Common to all figures: (a) view at source ↗
Figure 7.3
Figure 7.3. Figure 7.3: Schematic of VOC-net and VOC-lite training and testing procedures view at source ↗
Figure 7.4
Figure 7.4. Figure 7.4: Schematics of training and testing procedure of VOC-plus and VOC-certifire view at source ↗
Figure 7.5
Figure 7.5. Figure 7.5: Visual explanation of robustness radius R in smoothed classifiers. 7.3 Results and discussion To evaluate the performance of our models, we employed three key metrics: accuracy, F1- score, and precision. Detailed equations for these metrics are provided in the Supporting Information accompanying this paper. A concise overview of our findings is provided in view at source ↗
Figure 7.6
Figure 7.6. Figure 7.6: VOC-net confusion matrix. (a) Simulated data; (b) Experimental data view at source ↗
Figure 7.7
Figure 7.7. Figure 7.7: VOC-lite confusion matrix. (a) Simulated data; (b) Experimental data view at source ↗
Figure 7.8
Figure 7.8. Figure 7.8: VOC-plus confusion matrix. (a) Simulated data; (b) Experimental data view at source ↗
Figure 7.9
Figure 7.9. Figure 7.9: VOC-certifire confusion matrix. (a) Simulated data; (b) Experimental data view at source ↗
Figure 7.10
Figure 7.10. Figure 7.10: Certified accuracy analysis: (a) varying view at source ↗
read the original abstract

Laser absorption spectroscopy (LAS) is a well-established technique for non-intrusive measurement of gas species in combustion and atmospheric environments, but conventional methods struggle with multi-species mixtures under dynamic or interference-laden conditions. Overlapping spectral features, noise, and incomplete reference data limit reliability when unknown or weakly absorbing species are present. This dissertation develops diagnostics combining LAS with machine learning (ML) to address these limitations. Deep denoising autoencoders (DDAEs) are applied to shock-tube measurements during high-speed hydrocarbon pyrolysis, improving signal fidelity and detection limits for trace species. A structured unsupervised framework, HT-SIMNet, then mitigates interference from unknown species without full calibration data, using spectral augmentation and a Noise2Noise-inspired scheme to isolate species in reactive systems. Where reference spectra are unavailable, UnblindMix, an autoencoder-based blind source separation method, reconstructs concentrations and spectral signatures directly from mixture data, validated on mixtures of up to eight components. To recover weakly absorbing species masked by broader absorbers, a feature-engineering method based on first derivatives and convolutions selectively highlights minor species. Finally, VOC-certifire combines randomized smoothing with Voigt-based spectral perturbation to provide certifiable classification of volatile organic compounds under varying conditions. All techniques are experimentally validated and benchmarked. The integration of spectroscopic hardware with ML offers a path toward real-time, interference-resilient, reference-free gas detection for combustion science, environmental monitoring, and industrial safety.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript develops and experimentally validates several machine learning techniques integrated with laser absorption spectroscopy (LAS) for multi-species gas detection under challenging conditions. These include deep denoising autoencoders (DDAEs) applied to shock-tube hydrocarbon pyrolysis measurements, the HT-SIMNet framework for interference mitigation via spectral augmentation and Noise2Noise-inspired training, UnblindMix for autoencoder-based blind source separation on mixtures of up to eight components, derivative-and-convolution feature engineering to highlight weakly absorbing species, and VOC-certifire combining randomized smoothing with Voigt perturbations for certifiable classification. The work claims these enable real-time, reference-free, interference-resilient detection in combustion, environmental, and industrial settings.

Significance. If the reported experimental validations hold under broader conditions, the integration of ML with LAS hardware could meaningfully advance non-intrusive diagnostics where spectral overlap, noise, and incomplete references currently limit reliability. The experimental grounding on real shock-tube data and controlled mixtures is a positive aspect, as is the focus on practical challenges like unknown species and dynamic environments. However, the narrow scope of testing limits the immediate impact on the claimed application domains.

major comments (2)
  1. [Abstract] Abstract: the central claim that the techniques enable 'real-time, interference-resilient, reference-free gas detection' for 'unknown or weakly absorbing species' in 'complex and harsh environments' is not supported by the described validations, which are restricted to shock-tube pyrolysis experiments and controlled mixtures of at most eight components; no tests on novel species, alternate laser hardware, or more extreme dynamic conditions are reported.
  2. [Validation sections] Validation sections (implied by abstract descriptions of DDAE, HT-SIMNet, UnblindMix, and VOC-certifire): the assertion of 'experimentally validated and benchmarked' performance lacks any mention of quantitative metrics, error bars, data exclusion criteria, or cross-validation procedures, preventing assessment of whether post-hoc tuning or limited test conditions affect the robustness claims.
minor comments (1)
  1. [Abstract] Abstract: the phrasing 'All techniques are experimentally validated and benchmarked' is vague; specifying the exact benchmarks, comparison baselines, and performance metrics would improve clarity.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the thorough review and the recommendation for major revision. The comments highlight important points about the scope of our experimental validations and the clarity of quantitative reporting. We address each major comment below and outline revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the techniques enable 'real-time, interference-resilient, reference-free gas detection' for 'unknown or weakly absorbing species' in 'complex and harsh environments' is not supported by the described validations, which are restricted to shock-tube pyrolysis experiments and controlled mixtures of at most eight components; no tests on novel species, alternate laser hardware, or more extreme dynamic conditions are reported.

    Authors: The shock-tube pyrolysis experiments represent a harsh, high-speed, dynamic combustion environment with significant spectral interference and unknown species formation, while the controlled mixtures (up to eight components) directly test multi-species overlap and blind separation via UnblindMix. HT-SIMNet and the derivative-convolution method address interference and weak absorbers under these conditions. We agree, however, that the abstract overstates generality without qualifiers. We will revise the abstract to specify that the methods are 'experimentally demonstrated in shock-tube pyrolysis and controlled multi-component mixtures' and add a sentence on the potential for broader application while noting the current scope. revision: yes

  2. Referee: [Validation sections] Validation sections (implied by abstract descriptions of DDAE, HT-SIMNet, UnblindMix, and VOC-certifire): the assertion of 'experimentally validated and benchmarked' performance lacks any mention of quantitative metrics, error bars, data exclusion criteria, or cross-validation procedures, preventing assessment of whether post-hoc tuning or limited test conditions affect the robustness claims.

    Authors: Each technique section in the manuscript reports quantitative metrics (e.g., RMSE for reconstruction, detection limit improvements, classification accuracy with randomized smoothing bounds) and includes benchmark comparisons against conventional LAS methods, with error bars shown in figures. Data handling follows standard practices for the respective experiments. To improve accessibility and address the concern directly, we will insert a new summary paragraph or table in the validation overview that explicitly lists key metrics, data exclusion rules, and any cross-validation or train/test splits used across the methods. revision: yes

standing simulated objections not resolved
  • We cannot provide new experimental results on entirely novel species, alternate laser hardware platforms, or more extreme dynamic conditions, as these fall outside the scope of the completed study.

Circularity Check

0 steps flagged

No significant circularity; experimental validation of applied ML methods is self-contained

full rationale

The paper presents an applied dissertation that combines established laser absorption spectroscopy hardware with several machine-learning techniques (DDAEs, HT-SIMNet, UnblindMix, derivative feature engineering, and VOC-certifire). All central results are framed as experimental validations on shock-tube pyrolysis data and controlled multi-component mixtures rather than as first-principles derivations or predictions derived from fitted parameters. No equations, uniqueness theorems, or self-citations are invoked in a load-bearing manner that would reduce the reported performance metrics to quantities defined only in terms of the paper's own inputs. The work therefore contains no self-definitional, fitted-input-called-prediction, or self-citation-load-bearing steps.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claims rest on standard assumptions of ML model training (e.g., that training distributions capture test distributions) and spectroscopic line-shape models (Voigt profiles), plus several ML-specific hyperparameters whose values are not detailed in the abstract.

free parameters (2)
  • Autoencoder architecture hyperparameters
    Number of layers, latent dimension, and training schedule for DDAEs, HT-SIMNet, and UnblindMix are chosen to fit the spectroscopic data.
  • Spectral augmentation parameters
    Parameters controlling how reference spectra are perturbed to simulate unknown interferents.
axioms (2)
  • domain assumption Voigt line-shape profiles accurately represent absorption features under the tested conditions
    Invoked in the VOC-certifire component and implicit in all LAS processing.
  • domain assumption The training and test distributions of spectral interferences are sufficiently similar
    Required for the unsupervised and blind-separation methods to generalize.

pith-pipeline@v0.9.0 · 5558 in / 1502 out tokens · 20247 ms · 2026-05-09T18:30:34.320903+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

232 extracted references · 19 canonical work pages · 4 internal anchors

  1. [1]

    Drivers of improved pm2. 5 air quality in china from 2013 to 2017,

    Q. Zhang, Y . Zheng, D. Tong, M. Shao, S. Wang, Y . Zhang, X. Xu, J. Wang, H. He, W. Liuet al., “Drivers of improved pm2. 5 air quality in china from 2013 to 2017,” Proceedings of the National Academy of Sciences, vol. 116, no. 49, pp. 24 463– 24 469, 2019

    2013

  2. [2]

    WHO global air quality guidelines: particulate matter (PM2.5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide.https://www

    World Health Organization, “Who global air quality guidelines: Particulate matter (pm2.5 and pm 10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide,” https://www.who.int/publications/i/item/9789240034228, 2021, accessed: 2025-04- 21

    work page arXiv 2021

  3. [3]

    Who air quality guidelines 2021–aiming for healthier air for all: a joint statement by medical, public health, scientific societies and patient representative organisations,

    B. Hoffmann, H. Boogaard, A. de Nazelle, Z. J. Andersen, M. Abramson, M. Brauer, B. Brunekreef, F. Forastiere, W. Huang, H. Kanet al., “Who air quality guidelines 2021–aiming for healthier air for all: a joint statement by medical, public health, scientific societies and patient representative organisations,”International journal of public health, vol. 66...

    2021

  4. [4]

    National ambient air quality stan- dards (naaqs),

    United States Environmental Protection Agency, “National ambient air quality stan- dards (naaqs),” https://www.epa.gov/naaqs, 2021, accessed: 2025-04-21

    2021

  5. [5]

    Recent trends of volatile organic com- pounds in ambient air and its health impacts: A review,

    S. K. Chauhan, N. Saini, and V . B. Yadav, “Recent trends of volatile organic com- pounds in ambient air and its health impacts: A review,”Int. J. Technol. Res. Eng, vol. 1, no. 8, p. 667, 2014

    2014

  6. [6]

    Global concentrations of co2 and ch4 retrieved from gosat: First preliminary results,

    T. Yokota, Y . Yoshida, N. Eguchi, Y . Ota, T. Tanaka, H. Watanabe, and S. Maksyu- tov, “Global concentrations of co2 and ch4 retrieved from gosat: First preliminary results,”Sola, vol. 5, pp. 160–163, 2009

    2009

  7. [7]

    Comprehensive review on gas sensors: Unveiling recent developments and addressing challenges,

    S. Panda, S. Mehlawat, N. Dhariwal, A. Kumar, and A. Sanger, “Comprehensive review on gas sensors: Unveiling recent developments and addressing challenges,” Materials Science and Engineering: B, vol. 308, p. 117616, 2024

    2024

  8. [8]

    Chemical hazards and toxic substances,

    Occupational Safety and Health Administration (OSHA), “Chemical hazards and toxic substances,” https://www.osha.gov/chemical-hazards, 2021, accessed: 2025- 04-21

    2021

  9. [9]

    Confined spaces: Hazard atmospheres,

    ——, “Confined spaces: Hazard atmospheres,” https://www.osha.gov/ confined-spaces/hazard-atmosphere, 2021, accessed: 2025-04-21

    2021

  10. [10]

    Indoor air quality monitoring systems based on internet of things: A systematic review,

    J. Saini, M. Dutta, and G. Marques, “Indoor air quality monitoring systems based on internet of things: A systematic review,”International journal of environmental 135 research and public health, vol. 17, no. 14, p. 4942, 2020

    2020

  11. [11]

    Achieving better indoor air quality with iot systems for future buildings: Opportunities and challenges,

    X. Dai, W. Shang, J. Liu, M. Xue, and C. Wang, “Achieving better indoor air quality with iot systems for future buildings: Opportunities and challenges,”Science of The Total Environment, vol. 895, p. 164858, 2023

    2023

  12. [12]

    Development of a real-time moni- toring and detection indoor air quality system for intensive care unit and emergency department

    N. S. Baqer, H. Mohammed, A. Albahriet al., “Development of a real-time moni- toring and detection indoor air quality system for intensive care unit and emergency department.”Signa Vitae, vol. 19, no. 1, 2023

    2023

  13. [13]

    A calibration-free ammonia breath sensor using a quantum cascade laser with wms 2f/1f,

    K. Owen and A. Farooq, “A calibration-free ammonia breath sensor using a quantum cascade laser with wms 2f/1f,”Applied Physics B, vol. 116, pp. 371–383, 2014

    2014

  14. [14]

    Sniffing out the truth: clinical diagnosis using the elec- tronic nose,

    A. Pavlou and A. Turner, “Sniffing out the truth: clinical diagnosis using the elec- tronic nose,” 2000

    2000

  15. [15]

    Electronic nose technology in respiratory diseases,

    S. Dragonieri, G. Pennazza, P. Carratu, and O. Resta, “Electronic nose technology in respiratory diseases,”Lung, vol. 195, pp. 157–165, 2017

    2017

  16. [16]

    Nanomaterial-based sensors for detection of disease by volatile organic compounds,

    Y . Y . Broza and H. Haick, “Nanomaterial-based sensors for detection of disease by volatile organic compounds,”Nanomedicine, vol. 8, no. 5, pp. 785–806, 2013

    2013

  17. [17]

    Multiplexed mhz-rate mid-infrared laser absorption spectroscopy for simultaneous in-chamber co, co2, h2o, tempera- ture, and pressure in a rotating detonation rocket engine,

    N. M. Kuenning, A. P. Nair, A. R. Keller, N. Q. Minesi, E. Ozen, B. Bigler, J. Kriesel, J. W. Bennewitz, J. Burr, S. A. Danczyket al., “Multiplexed mhz-rate mid-infrared laser absorption spectroscopy for simultaneous in-chamber co, co2, h2o, tempera- ture, and pressure in a rotating detonation rocket engine,”Combustion and Flame, vol. 268, p. 113608, 2024

    2024

  18. [18]

    Ultrafast-laser-absorption spectroscopy in the mid-infrared for single-shot, calibration-free temperature and species measurements in low-and high-pressure combustion gases,

    R. J. Tancin and C. S. Goldenstein, “Ultrafast-laser-absorption spectroscopy in the mid-infrared for single-shot, calibration-free temperature and species measurements in low-and high-pressure combustion gases,”Optics Express, vol. 29, no. 19, pp. 30 140–30 154, 2021

    2021

  19. [19]

    De- tection of chemical warfare agents with a miniaturized high-performance drift tube ion mobility spectrometer using high-energetic photons for ionization,

    A. Ahrens, M. Allers, H. Bock, M. Hitzemann, A. Ficks, and S. Zimmermann, “De- tection of chemical warfare agents with a miniaturized high-performance drift tube ion mobility spectrometer using high-energetic photons for ionization,”Analytical chemistry, vol. 94, no. 44, pp. 15 440–15 447, 2022

    2022

  20. [20]

    Field-portable gas chro- matograph mass spectrometer (gc-ms) unit for semi-volatile compound analysis in groundwater,

    A. J. Bednar, A. L. Russell, T. Georgian, D. E. Splichal, C. A. Hayes, P. Tackett, L. V . Parker, R. A. Kirgan, M. Wells, D. Justeset al., “Field-portable gas chro- matograph mass spectrometer (gc-ms) unit for semi-volatile compound analysis in groundwater,” 2011

    2011

  21. [21]

    Fieldable mass spectrometry for forensic science, homeland security, and defense applications,

    K. Evans-Nguyen, A. R. Stelmack, P. C. Clowser, J. M. Holtz, and C. C. Mulligan, “Fieldable mass spectrometry for forensic science, homeland security, and defense applications,”Mass Spectrometry Reviews, vol. 40, no. 5, pp. 628–646, 2021

    2021

  22. [22]

    Gas phase multicomponent detection and analysis combining broad- band dual-frequency comb absorption spectroscopy and deep learning,

    L. Tian, J. Xia, A. A. Kolomenskii, H. A. Schuessler, F. Zhu, Y . Li, J. He, Q. Dong, 136 and S. Zhang, “Gas phase multicomponent detection and analysis combining broad- band dual-frequency comb absorption spectroscopy and deep learning,”Communi- cations Engineering, vol. 2, no. 1, p. 54, 2023

    2023

  23. [23]

    Real-time gas mass spectroscopy by multivariate analysis,

    L. Franceschelli, C. Ciricugno, M. Di Lorenzo, A. Romani, A. Berardinelli, M. Tartagni, and R. Correale, “Real-time gas mass spectroscopy by multivariate analysis,”Scientific Reports, vol. 13, no. 1, p. 6059, 2023

    2023

  24. [24]

    O. D. Sparkman, Z. Penton, and F. G. Kitson,Gas chromatography and mass spec- trometry: a practical guide. Academic press, 2011

    2011

  25. [25]

    Miniature and fieldable mass spectrometers: recent advances,

    D. T. Snyder, C. J. Pulliam, Z. Ouyang, and R. G. Cooks, “Miniature and fieldable mass spectrometers: recent advances,”Analytical chemistry, vol. 88, no. 1, pp. 2–29, 2016

    2016

  26. [26]

    High-resolution mass spectrometers,

    A. G. Marshall and C. L. Hendrickson, “High-resolution mass spectrometers,”Annu. Rev. Anal. Chem., vol. 1, no. 1, pp. 579–599, 2008

    2008

  27. [27]

    Time-of-flight mass spectrometry,

    R. J. Cotter, “Time-of-flight mass spectrometry,”Encyclopedia of Genetics, Ge- nomics, Proteomics and Bioinformatics, 2004

    2004

  28. [28]

    Fourier transform infrared spectrometry,

    P. R. Griffiths, “Fourier transform infrared spectrometry,”Science, vol. 222, no. 4621, pp. 297–302, 1983

    1983

  29. [29]

    Infrared laser- absorption sensing for combustion gases,

    C. S. Goldenstein, R. M. Spearrin, J. B. Jeffries, and R. K. Hanson, “Infrared laser- absorption sensing for combustion gases,”Progress in Energy and Combustion Sci- ence, vol. 60, pp. 132–176, 2017

    2017

  30. [30]

    Laser sensors for energy systems and process industries: Perspectives and directions,

    A. Farooq, A. B. Alquaity, M. Raza, E. F. Nasir, S. Yao, and W. Ren, “Laser sensors for energy systems and process industries: Perspectives and directions,”Progress in Energy and Combustion Science, vol. 91, p. 100997, 2022

    2022

  31. [31]

    Laser absorption spectroscopy for combustion diagnosis in re- active flows: A review,

    C. Liu and L. Xu, “Laser absorption spectroscopy for combustion diagnosis in re- active flows: A review,”Applied Spectroscopy Reviews, vol. 54, no. 1, pp. 1–44, 2019

    2019

  32. [32]

    A multi-wavelength speciation framework for high-temperature hy- drocarbon pyrolysis,

    N. H. Pinkowski, Y . Ding, S. E. Johnson, Y . Wang, T. C. Parise, D. F. Davidson, and R. K. Hanson, “A multi-wavelength speciation framework for high-temperature hy- drocarbon pyrolysis,”Journal of Quantitative Spectroscopy and Radiative Transfer, vol. 225, pp. 180–205, 2019

    2019

  33. [34]

    P. F. Bernath,Spectra of atoms and molecules. Oxford university press, 2020

    2020

  34. [35]

    Selectivity in trace gas sensing: recent developments, challenges, and future perspectives,

    P. Barik and M. Pradhan, “Selectivity in trace gas sensing: recent developments, challenges, and future perspectives,”Analyst, vol. 147, no. 6, pp. 1024–1054, 2022

    2022

  35. [36]

    Exomol: molecular line lists for exoplanet and 137 other atmospheres,

    J. Tennyson and S. N. Yurchenko, “Exomol: molecular line lists for exoplanet and 137 other atmospheres,”Monthly Notices of the Royal Astronomical Society, vol. 425, no. 1, pp. 21–33, 2012

    2012

  36. [37]

    Gas-phase databases for quantitative infrared spectroscopy,

    S. W. Sharpe, T. J. Johnson, R. L. Sams, P. M. Chu, G. C. Rhoderick, and P. A. John- son, “Gas-phase databases for quantitative infrared spectroscopy,”Appl. Spectrosc, vol. 58, no. 12, pp. 1452–1461, 2004

    2004

  37. [38]

    Artificial intelligence in gas sensing: A re- view,

    M. Chowdhury and M. Oehlschlaeger, “Artificial intelligence in gas sensing: A re- view,”ACS sensors, 2025

    2025

  38. [39]

    Mixtures recomposition by neural nets: a multidisciplinary overview,

    A. Nicolle, S. Deng, M. Ihme, N. Kuzhagaliyeva, E. A. Ibrahim, and A. Farooq, “Mixtures recomposition by neural nets: a multidisciplinary overview,”Journal of Chemical Information and Modeling, vol. 64, no. 3, pp. 597–620, 2024

    2024

  39. [40]

    Combustion machine learning: Princi- ples, progress and prospects,

    M. Ihme, W. T. Chung, and A. A. Mishra, “Combustion machine learning: Princi- ples, progress and prospects,”Progress in Energy and Combustion Science, vol. 91, p. 101010, 2022

    2022

  40. [41]

    Artificial intelligence as a catalyst for combustion sci- ence and engineering,

    M. Ihme and W. T. Chung, “Artificial intelligence as a catalyst for combustion sci- ence and engineering,”Proceedings of the Combustion Institute, vol. 40, no. 1-4, p. 105730, 2024

    2024

  41. [42]

    Predicting infrared spectra with message passing neural networks,

    C. McGill, M. Forsuelo, Y . Guan, and W. H. Green, “Predicting infrared spectra with message passing neural networks,”Journal of chemical information and modeling, vol. 61, no. 6, pp. 2594–2609, 2021

    2021

  42. [43]

    Octane prediction from infrared spectroscopic data,

    E. Al Ibrahim and A. Farooq, “Octane prediction from infrared spectroscopic data,” Energy & Fuels, vol. 34, no. 1, pp. 817–826, 2019

    2019

  43. [44]

    Chemprop: a machine learning package for chem- ical property prediction,

    E. Heid, K. P. Greenman, Y . Chung, S.-C. Li, D. E. Graff, F. H. Vermeire, H. Wu, W. H. Green, and C. J. McGill, “Chemprop: a machine learning package for chem- ical property prediction,”Journal of Chemical Information and Modeling, vol. 64, no. 1, pp. 9–17, 2023

    2023

  44. [45]

    Prediction of the derived cetane number and car- bon/hydrogen ratio from infrared spectroscopic data,

    E. Al Ibrahim and A. Farooq, “Prediction of the derived cetane number and car- bon/hydrogen ratio from infrared spectroscopic data,”Energy & Fuels, vol. 35, no. 9, pp. 8141–8152, 2021

    2021

  45. [46]

    Chemical gas sensors: Recent developments, chal- lenges, and the potential of machine learning—a review,

    U. Yaqoob and M. I. Younis, “Chemical gas sensors: Recent developments, chal- lenges, and the potential of machine learning—a review,”Sensors, vol. 21, no. 8, p. 2877, 2021

    2021

  46. [47]

    Selective btex measurements using deep neural networks,

    M. Mhanna, M. Sy, and A. Farooq, “Selective btex measurements using deep neural networks,” inCLEO: Science and Innovations. Optica Publishing Group, 2021, pp. JTh3A–49

    2021

  47. [48]

    Deep learning for gas sensing via infrared spectroscopy,

    M. A. Z. Chowdhury and M. A. Oehlschlaeger, “Deep learning for gas sensing via infrared spectroscopy,”Sensors, vol. 24, no. 6, p. 1873, 2024. 138

    2024

  48. [49]

    Selective btex detection using laser absorption spectroscopy in the ch bending mode region,

    A. Elkhazraji, M. Sy, M. Mhanna, J. Aldhawyan, M. K. Shakfa, and A. Farooq, “Selective btex detection using laser absorption spectroscopy in the ch bending mode region,”Experimental Thermal and Fluid Science, vol. 151, p. 111090, 2024

    2024

  49. [50]

    Deep neural networks for si- multaneous btex sensing at high temperatures,

    M. Mhanna, M. Sy, A. Elkhazraji, and A. Farooq, “Deep neural networks for si- multaneous btex sensing at high temperatures,”Optics Express, vol. 30, no. 21, pp. 38 550–38 563, 2022

    2022

  50. [51]

    Laser-based selective btex sensing using deep neural networks,

    M. Mhanna, M. Sy, A. Arfaj, J. Llamas, and A. Farooq, “Laser-based selective btex sensing using deep neural networks,”Optics Letters, vol. 47, no. 13, pp. 3247–3250, 2022

    2022

  51. [52]

    Multi-speciation in shock tube kinetics using deep neural networks and cavity-enhanced absorption spectroscopy,

    M. Mhanna, M. Sy, A. Elkhazraji, and A. Farooq, “Multi-speciation in shock tube kinetics using deep neural networks and cavity-enhanced absorption spectroscopy,” Proceedings of the Combustion Institute, vol. 40, no. 1-4, p. 105733, 2024

    2024

  52. [53]

    Blind source separation method based on neu- ral network with bias term and maximum likelihood estimation criterion,

    S. Liu, B. Wang, and L. Zhang, “Blind source separation method based on neu- ral network with bias term and maximum likelihood estimation criterion,”Sensors, vol. 21, no. 3, p. 973, 2021

    2021

  53. [54]

    Unsupervised blind source separation with vari- ational auto-encoders,

    J. Neri, R. Badeau, and P. Depalle, “Unsupervised blind source separation with vari- ational auto-encoders,” inProceedings of the 29th European Signal Processing Con- ference (EUSIPCO), Dublin, Ireland, 2021

    2021

  54. [55]

    Human internal state estimation as blind source separation using a dynamic auto-encoder,

    C. A. Steed and N. Kim, “Human internal state estimation as blind source separation using a dynamic auto-encoder,” in2023 15th International Conference on Advanced Computational Intelligence (ICACI). IEEE, 2023, pp. 1–6

    2023

  55. [56]

    Leveraging deep learning for optimal methane gas detection: Residual network filter assisted direct absorption spectroscopy,

    R. Xu, L. Tian, J. Xia, F. Zhao, K. Guo, Z. Liang, and S. Zhang, “Leveraging deep learning for optimal methane gas detection: Residual network filter assisted direct absorption spectroscopy,”Sensors and Actuators A: Physical, vol. 369, p. 115195, 2024

    2024

  56. [57]

    The beer-lambert law,

    D. F. Swinehart, “The beer-lambert law,”J. Chem. Educ., vol. 39, no. 7, p. 333, 1962

    1962

  57. [58]

    Deep learning for near-infrared spectral data modelling: Hypes and benefits,

    P. Mishra, D. Passos, F. Marini, J. Xu, J. Amigo, A. Gowen, J. Jansen, A. Biancolillo, J. Roger, D. Rutledge, and A. Nordon, “Deep learning for near-infrared spectral data modelling: Hypes and benefits,”TrAC Trends in Analytical Chemistry, vol. 157, p. 116804, 2022

    2022

  58. [59]

    A scoping review of infrared spectroscopy and machine learning methods for head and neck precancer and cancer diagnosis and prognosis,

    S. Alajaji, R. Sabzian, Y . Wang, A. Sultan, and R. Wang, “A scoping review of infrared spectroscopy and machine learning methods for head and neck precancer and cancer diagnosis and prognosis,”Cancers, vol. 17, no. 5, p. 796, 2025

    2025

  59. [60]

    Supervised learning,

    P. Cunningham, M. Cord, and S. Delany, “Supervised learning,” inMachine Learning Techniques for Multimedia, ser. Cognitive Technologies, M. Cord and P. Cunningham, Eds. Springer, Berlin, Heidelberg, 2008, pp. 21–49. [Online]. 139 Available: https://doi.org/10.1007/978-3-540-75171-7 2

    work page doi:10.1007/978-3-540-75171-7 2008

  60. [61]

    Explainable artificial intelligence for spectroscopy data: a review,

    J. Contreras and T. Bocklitz, “Explainable artificial intelligence for spectroscopy data: a review,”Pfl ¨ugers Archiv-European Journal of Physiology, vol. 477, no. 4, pp. 603–615, 2025

    2025

  61. [62]

    Unsupervised learning,

    Z. Ghahramani, “Unsupervised learning,” inAdvanced Lectures on Machine Learning, ser. Lecture Notes in Computer Science, O. Bousquet, U. von Luxburg, and G. R ¨atsch, Eds. Springer, Berlin, Heidelberg, 2004, vol. 3176, pp. 72–112. [Online]. Available: https://doi.org/10.1007/978-3-540-28650-9 5

    work page doi:10.1007/978-3-540-28650-9 2004

  62. [63]

    Blind and endmember guided autoencoder model for unmixing the absorbance spectra of phytoplankton pigments,

    P. Naik, I. P ¨ol¨onen, and P. Salmi, “Blind and endmember guided autoencoder model for unmixing the absorbance spectra of phytoplankton pigments,” Scientific Reports, vol. 15, p. 13157, 2025. [Online]. Available: https: //doi.org/10.1038/s41598-025-96023-5

    work page doi:10.1038/s41598-025-96023-5 2025

  63. [64]

    Calibration of spectra in presence of non-stationary background using unsupervised physics-informed deep learning,

    A. Puleio, R. Rossi, and P. Gaudio, “Calibration of spectra in presence of non-stationary background using unsupervised physics-informed deep learning,” Scientific Reports, vol. 13, no. 1, p. 2156, 2023. [Online]. Available: https://doi.org/10.1038/s41598-023-29371-9

    work page doi:10.1038/s41598-023-29371-9 2023

  64. [65]

    LeCun, Y

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning,”Nature, vol. 521, no. 7553, pp. 436–444, 2015. [Online]. Available: https://doi.org/10.1038/nature14539

    work page doi:10.1038/nature14539 2015

  65. [66]

    Test- ing the limits of a portable nir spectrometer: content uniformity of complex powder mixtures followed by calibration transfer for in-line blend monitoring,

    T. Casian, A. Gavan, S. Iurian, A. Porfire, V . Toma, R. Stiufiuc, and I. Tomuta, “Test- ing the limits of a portable nir spectrometer: content uniformity of complex powder mixtures followed by calibration transfer for in-line blend monitoring,”Molecules, vol. 26, no. 4, p. 1129, 2021

    2021

  66. [67]

    Linear and non-linear modelling methods for a gas sensor array developed for process control applications,

    R. Lakhmi, M. Fischer, Q. Darves-Blanc, R. Alrammouz, M. Rieu, and J.-P. Viri- celle, “Linear and non-linear modelling methods for a gas sensor array developed for process control applications,”Sensors, vol. 24, no. 11, p. 3499, 2024

    2024

  67. [68]

    Tdacnn: Target-domain-free domain adaptation convolutional neural network for drift com- pensation in gas sensors,

    Y . Zhang, S. Xiang, Z. Wang, X. Peng, Y . Tian, S. Duan, and J. Yan, “Tdacnn: Target-domain-free domain adaptation convolutional neural network for drift com- pensation in gas sensors,”Sensors and Actuators B: Chemical, vol. 361, p. 131739, 2022

    2022

  68. [69]

    Co-training neural network-based in- frared sensor array for natural gas monitoring,

    J. Wang, S. Lian, B. Lei, B. Li, and S. Lei, “Co-training neural network-based in- frared sensor array for natural gas monitoring,”Sensors and Actuators A: Physical, vol. 335, p. 113392, 2022

    2022

  69. [71]

    Analysis of gas mixtures with broadband dual frequency comb spec- 140 troscopy and unsupervised learning neural network,

    L. Tianet al., “Analysis of gas mixtures with broadband dual frequency comb spec- 140 troscopy and unsupervised learning neural network,”Communications Engineering, vol. 2, p. 105, 2023

    2023

  70. [72]

    Tdacnn: Target-domain-free domain adaptation convolutional neural network for drift com- pensation in gas sensors,

    Y . Zhang, S. Xiang, Z. Wang, X. Peng, Y . Tian, S. Duan, and J. Yan, “Tdacnn: Target-domain-free domain adaptation convolutional neural network for drift com- pensation in gas sensors,”ArXiv preprint arXiv:2107.08274, 2021

    work page arXiv 2021

  71. [73]

    Denoising and baseline correction methods for ra- man spectroscopy based on convolutional autoencoder: A unified solution,

    M. Han, Y . Dang, and J. Han, “Denoising and baseline correction methods for ra- man spectroscopy based on convolutional autoencoder: A unified solution,”Sensors, vol. 24, no. 10, p. 3161, 2024

    2024

  72. [74]

    Deep learning-based denoising for fast time-resolved flame emission spectroscopy in high-pressure combustion environ- ment,

    T. Yoon, S. W. Kim, H. Byunet al., “Deep learning-based denoising for fast time-resolved flame emission spectroscopy in high-pressure combustion environ- ment,”arXiv preprint, 2022

    2022

  73. [75]

    Convolution net- work with custom loss function for the denoising of low snr raman spectra,

    S. Barton, S. Alakkari, K. O’Dwyer, T. Ward, and B. Hennelly, “Convolution net- work with custom loss function for the denoising of low snr raman spectra,”Sensors, vol. 21, no. 14, p. 4823, 2021

    2021

  74. [76]

    Convolutional neural network and guided filtering for sar image denoising,

    S. Liu, T. Liu, L. Gaoet al., “Convolutional neural network and guided filtering for sar image denoising,”Remote Sensing, vol. 11, no. 6, p. 702, 2019

    2019

  75. [77]

    Laser absorption spectroscopy based on dual-convolutional neu- ral network algorithms for multiple trace gases analysis,

    F. n. w. Renet al., “Laser absorption spectroscopy based on dual-convolutional neu- ral network algorithms for multiple trace gases analysis,”arXiv preprint, 2024

    2024

  76. [78]

    Principal component analysis: a review and recent de- velopments,

    I. T. Jolliffe and J. Cadima, “Principal component analysis: a review and recent de- velopments,”Philosophical transactions of the royal society A: Mathematical, Phys- ical and Engineering Sciences, vol. 374, no. 2065, p. 20150202, 2016

    2065

  77. [79]

    Independent component analysis: algorithms and appli- cations,

    A. Hyv ¨arinen and E. Oja, “Independent component analysis: algorithms and appli- cations,”Neural Netw., vol. 13, no. 4-5, pp. 411–430, 2000

    2000

  78. [80]

    Learning the parts of objects by non-negative matrix factorization,

    D. D. Lee and H. S. Seung, “Learning the parts of objects by non-negative matrix factorization,”nature, vol. 401, no. 6755, pp. 788–791, 1999

    1999

  79. [81]

    Deep learn- ing for visualization and novelty detection in large x-ray diffraction datasets,

    L. Banko, P. M. Maffettone, D. Naujoks, D. Olds, and A. Ludwig, “Deep learn- ing for visualization and novelty detection in large x-ray diffraction datasets,”Npj Computational Materials, vol. 7, no. 1, p. 104, 2021

    2021

  80. [82]

    Deep attractor network for single-microphone speaker separation,

    Z. Chen, Y . Luo, and N. Mesgarani, “Deep attractor network for single-microphone speaker separation,” in2017 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2017, pp. 246–250

    2017

Showing first 80 references.