Gaussian process based modeling and experimental design for sensor calibration in drifting environments.

Area of Science:

• Sensor technology
• Environmental monitoring
• Statistical modeling

Background:

• Accurate sensor calibration is challenging due to environmental drift.
• Sensor response depends on analyte concentration and environmental factors like temperature and humidity.
• Existing methods struggle with dynamic environmental conditions.

Purpose of the Study:

• To develop an efficient calibration procedure for sensors in drifting environments.
• To utilize Gaussian Process (GP) modeling for capturing complex sensor response relationships.
• To enable uncertainty quantification for estimated analyte concentrations.

Main Methods:

• Developed a Gaussian Process (GP)-based calibration model.
• Integrated GP's inference capabilities with an experimental design for efficient data sampling.
• Employed batch sequential data acquisition for calibration.

Main Results:

• The GP-based procedure effectively models nonlinear relationships between sensor response and exposure factors.
• The method provides valid statistical inference for uncertainty quantification.
• Demonstrated superior effectiveness and efficiency compared to traditional calibration methods on a simulated chemiresistor sensor.

Conclusions:

• The proposed GP-based calibration procedure offers an efficient solution for sensors in drifting environments.
• This approach enhances the accuracy of analyte concentration estimation and provides reliable uncertainty measures.
• The integrated modeling and experimental design method advances sensor calibration techniques.