Scholarlink

In this article authors are going to represent a Photonic Crystal Gas Sensor device using Machine Learning (ML) approach. Photonic crystals are engineered materials with periodic optical structures, characterized by the formation of photonic band gaps (PBGs) that control the propagation of light. Any disturbance in their periodicity, such as the introduction of a defect layer, leads to the emergence of localized states within the PBG. These states are highly sensitive to variations in the refractive index of the defect layer, a property exploited in gas detection. By harnessing this principle, PhC-based sensors have been utilized for detecting gases like argon, helium, and ozone, among others. One-dimensional (1D) PhCs, in particular, have shown immense potential due to their structural simplicity and efficiency in detecting minute changes in refractive index. Despite their potential, the manual design and tuning of PhC structures for diverse gases remain a time-intensive process. This has paved the way for integrating machine learning (ML) techniques with PhC-based sensors, ushering in a new era of smart sensing solutions. Machine learning algorithms, particularly Support Vector Machines (SVM) and Random Forest (RF), offer the capability to analyze large datasets, identify patterns, and enhance prediction accuracy. These algorithms, when combined with PhCs, enable rapid classification and density prediction of gases under varying environmental conditions, thereby overcoming the limitations of traditional methods. This study introduces a novel approach to gas sensing by combining the advantages of 1D PhC structures with machine learning algorithms for detecting SF6, CH4, and CO2. Through extensive simulations, datasets were generated across a wide range of temperatures and pressures. These datasets were used to train and optimize the SVM and RF classifiers, achieving classification accuracies as high as 97.01% with SVM. The research emphasizes the role of hyperparameter tuning in optimizing the performance of machine learning models, demonstrating the consistent reliability of SVM over RF for real-time applications.

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

Design and Analysis of a Photonic Crystal Gas Sensor Using Machine Learning Approach

  • 章节

  • 发表日期：

    2026年

查看全文
  • Kanishka Majumder,
  • Jayabrata Goswami

摘要

In this article authors are going to represent a Photonic Crystal Gas Sensor device using Machine Learning (ML) approach. Photonic crystals are engineered materials with periodic optical structures, characterized by the formation of photonic band gaps (PBGs) that control the propagation of light. Any disturbance in their periodicity, such as the introduction of a defect layer, leads to the emergence of localized states within the PBG. These states are highly sensitive to variations in the refractive index of the defect layer, a property exploited in gas detection. By harnessing this principle, PhC-based sensors have been utilized for detecting gases like argon, helium, and ozone, among others. One-dimensional (1D) PhCs, in particular, have shown immense potential due to their structural simplicity and efficiency in detecting minute changes in refractive index. Despite their potential, the manual design and tuning of PhC structures for diverse gases remain a time-intensive process. This has paved the way for integrating machine learning (ML) techniques with PhC-based sensors, ushering in a new era of smart sensing solutions. Machine learning algorithms, particularly Support Vector Machines (SVM) and Random Forest (RF), offer the capability to analyze large datasets, identify patterns, and enhance prediction accuracy. These algorithms, when combined with PhCs, enable rapid classification and density prediction of gases under varying environmental conditions, thereby overcoming the limitations of traditional methods. This study introduces a novel approach to gas sensing by combining the advantages of 1D PhC structures with machine learning algorithms for detecting SF6, CH4, and CO2. Through extensive simulations, datasets were generated across a wide range of temperatures and pressures. These datasets were used to train and optimize the SVM and RF classifiers, achieving classification accuracies as high as 97.01% with SVM. The research emphasizes the role of hyperparameter tuning in optimizing the performance of machine learning models, demonstrating the consistent reliability of SVM over RF for real-time applications.