Natural Gas Analysis in Hazardous Locations Using Raman Spectroscopy and Chemometric Modeling

Significance of the topic

Accurate compositional analysis of natural gas is essential for custody transfer, process control, regulatory compliance, safety and emissions management. Measuring gas composition directly in hazardous environments (ATEX Zone 0 / IECEx / Class 1 Division 1) reduces sampling risk and latency compared with conventional sampling loops and lab analysis. Raman spectroscopy, when implemented using intrinsically safe hardware and combined with advanced chemometric models, can deliver rapid, in-situ quantification of major and minor hydrocarbon species and common contaminants in pipeline- or process-line conditions.

Objectives and overview of the study

This application note demonstrates the viability of the Thermo Scientific MarqMetrix All-In-One X Process Raman Analyzer coupled with a MarqMetrix FlowCell Sampling Optic for direct natural gas analysis in hazardous locations. Objectives were to acquire high-quality Raman spectra from six certified natural gas standards under ATEX-compliant conditions and to develop robust Partial Least Squares (PLS) chemometric models to quantify key components (methane, C2–C6 hydrocarbons, N2, CO2).

Methodology and data acquisition

• Sampling and conditions: Six natural gas cylinder standards (compositions spanning methane-dominated mixtures with variable ethane, propane, butanes, pentanes, nitrogen and CO2) were connected via a gas sampling rig. Pressure was held constant at 100 psi. Each spectrum was an average of six acquisitions; acquisition time per replicate was 1 minute.
• Instrumentation: ATEX-compliant 35 mW Raman laser integrated in the MarqMetrix All-In-One X analyzer with MarqMetrix FlowCell (certified for ATEX Zone 0 / IECEx / Class 1 Division 1). Spectra were imported into Solo 9.5 (Eigenvector) for chemometric analysis.
• Spectral treatment: Key Raman regions were selected per analyte to reduce cross-covariance between hydrocarbons (examples: methane 2880–2960 cm-1; ethane 975–1050 cm-1; propane 850–950 cm-1; butanes/pentanes 370–700 and 800–875 cm-1; nitrogen 2320–2360 cm-1; CO2 1250–1500 cm-1).
• Chemometrics and validation: PLS regression models were built with targeted variable selection and preprocessing. Model complexity (number of latent variables) was optimized for each analyte. Due to a limited sample set, model performance was evaluated by RMSECV using Venetian blinds cross-validation holding all replicates together; 95% CI for RMSECV reported as RMSECV × 1.96.

Used instrumentation

• Thermo Scientific MarqMetrix All-In-One X Process Raman Analyzer (intrinsically safe configuration)
• Thermo Scientific MarqMetrix FlowCell Sampling Optic (ATEX Zone 0 / IECEx / Class 1 Division 1 certified)
• ATEX-compliant 35 mW laser, sampling rig maintained at 100 psi
• Solo 9.5 chemometrics software (Eigenvector Research) for PLS model development and validation

Main results and discussion

• Spectral quality: Raman spectra exhibited well-resolved, low-fluorescence peaks for major components, facilitating selective variable selection and robust modeling.
• Model performance highlights (selected metrics):
  • Methane (range 72.7–90 mol%): RMSECV 1.91 mol%, R² = 0.993
  • Ethane (2.5–10 mol%): RMSECV 0.075 mol%, R² = 0.999
  • Propane (1–4.6 mol%): RMSECV 0.095 mol%, R² = 0.995
  • Carbon dioxide (0–10.4 mol%): RMSECV 0.182 mol%, R² = 0.999
  • Nitrogen (0.2–9 mol%): RMSECV 0.575 mol%, R² = 0.968
  • i‑Butane / n‑Butane and i‑Pentane / n‑Pentane: higher relative uncertainty and lower R² (but still useful), reflecting lower concentrations and weaker Raman intensities; wider variable ranges were used to capture signal.
• Model characteristics: Appropriate spectral region selection reduced covariance effects among hydrocarbons, improving model independence. Latent variables per model were optimized (typical LV counts ranged from 2 to 5). The plots of predicted vs measured values showed strong linearity and minimal bias for major species.
• Limitations: Dataset size was limited (six standards with replicates), so reported RMSECV-based performance is indicative but would benefit from expanded calibration sets, external validation, and sampling under broader field conditions (temperature/pressure/mixtures) for transferability.

Benefits and practical applications

• Intrinsic safety enables deployment directly in hazardous zones, eliminating the need for purge/sample loops and reducing risk and maintenance.
• Near-real-time quantification of major hydrocarbons, CO2 and N2 supports custody-transfer checks, process control, leak detection and compositional monitoring for combustion optimization.
• Non-invasive, in-situ Raman measurement reduces sample handling artifacts and provides faster decisions for plant operations and safety systems.
• High selectivity to vibrational signatures facilitates discrimination of C1–C6 hydrocarbons even in mixed streams; chemometric modeling improves sensitivity to low-level species.

Future trends and potential uses

• Expanded calibration libraries: Building larger, more diverse training sets (varying temperature, pressure, matrix composition) will improve model robustness and enable transfer between sites.
• Model transfer and calibration maintenance: Techniques such as standardization, transfer learning and adaptive calibration could reduce recalibration effort in the field.
• Integration with process control and edge analytics: Embedding chemometric models and drift correction on instrument firmware or edge devices will enable faster closed-loop control and anomaly detection.
• Combining Raman with other sensors: Hybrid approaches (e.g., Raman + TDLAS, GC, or NIR) can extend dynamic range and detection limits for trace impurities and complex mixtures.
• AI-driven preprocessing and variable selection: Advanced algorithms can automate region selection and compensate for baseline/drift, yielding improved sensitivity for low-intensity species (butanes, pentanes).
• Wider deployment scenarios: Mobile and distributed Raman analyzers for pipeline monitoring, compressor station surveillance, LNG quality checks and emissions monitoring in hazardous areas.

Conclusion

The study demonstrates that an intrinsically safe Raman analyzer (MarqMetrix All-In-One X) with a certified flow cell and focused chemometric modeling can deliver accurate, rapid compositional analysis of natural gas mixtures in hazardous locations. Careful spectral region selection and PLS optimization produced high linearity and low prediction error for major species (methane, ethane, propane, CO2) while revealing the expected challenges for low-concentration heavier hydrocarbons. The approach reduces sampling complexity and enables near real-time decision support for safety, custody, and process control, though broader calibration sets and field validation are recommended to strengthen model transfer and long-term robustness.

Reference

• Thermo Scientific MarqMetrix All-In-One X Process Raman Analyzer application note: Natural Gas Analysis in Hazardous Locations Using Raman Spectroscopy and Chemometric Modeling (Thermo Fisher Scientific, MCS-AN1350-EN 3/25). 2025.