Fault Detection Using the Clustering-kNN Rule for Gas Sensor Arrays

Area of Science:

• Sensor technology
• Machine learning
• Signal processing

Background:

• Gas sensor arrays are crucial for environmental monitoring but susceptible to faults.
• Traditional k-nearest neighbour (kNN) fault detection methods are computationally intensive, hindering real-time applications.
• Existing methods struggle with large datasets and non-linear data, impacting monitoring reliability.

Purpose of the Study:

• To develop an efficient and accurate fault detection method for gas sensor arrays.
• To address the computational limitations of traditional kNN fault detection.
• To enable reliable real-time monitoring of gas sensor systems.

Main Methods:

• A novel clustering-kNN rule integrating landmark-based spectral clustering (LSC) was developed.
• LSC algorithm efficiently partitions the training dataset into clusters with low computational complexity.
• The kNN rule is applied only to the nearest cluster, significantly reducing computational load per test sample.

Main Results:

• The clustering-kNN rule demonstrated enhanced accuracy and efficiency in numerical simulations and real-world gas sensor experiments.
• Performance was validated across linear and non-linear data models and diverse fault types.
• Significant improvements in fault detection speed and reliability were observed compared to traditional methods.

Conclusions:

• The proposed clustering-kNN rule offers a robust solution for accurate and efficient fault detection in gas sensor arrays.
• This approach facilitates reliable real-time monitoring, overcoming the limitations of conventional techniques.
• The method shows great promise for enhancing the dependability of gas sensing systems in various applications.