Scholarlink

Semiconductor metal oxide (MOS) gas sensors employing metal oxide semiconductor materials such as WO₃, SnO₂, ZnO, TiO₂, and V₂O₅… are widely used to detect the presence of hazardous gases in the environment, including NO₂, NH₃, and H₂S. This paper proposes the development of an integrated framework combining statistical learning and Bayesian inference through Gaussian Process Regression (GPR) to optimize and assess the reliability of H₂S sensors fabricated from WO₃/rGO nanocomposites. Time-resolved resistance data were collected from five sensor samples under various operating temperatures (200–450 ℃) and H₂S concentrations (1–20 ppm). After preprocessing and feature extraction, techniques such as Principal Component Analysis (PCA), Naive Bayes classification, and GPR were applied to identify the optimal sensor samples and predict gas concentrations. The results demonstrate that GPR achieves a low prediction error (RMSE < 1.5 ppm) while simultaneously providing probabilistic confidence intervals that reflect predictive uncertainty. This study highlights the potential of integrating nano-scale sensing materials with advanced statistical learning approaches for the development of intelligent and robust gas sensing systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Gaussian Process Bayesian Regression and PCA Framework for Reliability Evaluation of H₂S Gas Sensors Based on WO₃/RGO Nanocomposites

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Dao Sy Nhien,
  • Do Quang Dat,
  • Nguyen Thi Thu Ha,
  • Bui Thi Tuyet,
  • Hoang Cao Minh

摘要

Semiconductor metal oxide (MOS) gas sensors employing metal oxide semiconductor materials such as WO₃, SnO₂, ZnO, TiO₂, and V₂O₅… are widely used to detect the presence of hazardous gases in the environment, including NO₂, NH₃, and H₂S. This paper proposes the development of an integrated framework combining statistical learning and Bayesian inference through Gaussian Process Regression (GPR) to optimize and assess the reliability of H₂S sensors fabricated from WO₃/rGO nanocomposites. Time-resolved resistance data were collected from five sensor samples under various operating temperatures (200–450 ℃) and H₂S concentrations (1–20 ppm). After preprocessing and feature extraction, techniques such as Principal Component Analysis (PCA), Naive Bayes classification, and GPR were applied to identify the optimal sensor samples and predict gas concentrations. The results demonstrate that GPR achieves a low prediction error (RMSE < 1.5 ppm) while simultaneously providing probabilistic confidence intervals that reflect predictive uncertainty. This study highlights the potential of integrating nano-scale sensing materials with advanced statistical learning approaches for the development of intelligent and robust gas sensing systems.