Dual-optimized adaptive Kalman filtering algorithm based on BP neural network and variance compensation for laser...

Area of Science:

• Laser Spectroscopy
• Signal Processing
• Analytical Chemistry

Background:

• Accurate trace gas detection is crucial for environmental monitoring and medical diagnostics.
• Traditional methods often struggle with sensitivity and parameter variations in dynamic systems.
• Adaptive Kalman filtering offers potential but requires robust parameter optimization.

Purpose of the Study:

• To develop a highly sensitive trace gas detection algorithm for laser spectroscopy.
• To improve Kalman filter (KF) performance using neural networks and variance compensation.
• To validate the algorithm's efficacy in a real-world gas sensing application.

Main Methods:

• Developed a dual-optimized adaptive Kalman filtering (DO-AKF) algorithm.
• Integrated a back propagation (BP) neural network for KF parameter optimization.
• Incorporated variance compensation to manage dynamic system parameter variations.
• Compared DO-AKF against traditional multi-signal average, extended KF, unscented KF, BP-KF, and VC-KF.

Main Results:

• The DO-AKF algorithm demonstrated superior performance over all compared methods.
• Applied to a QCL-based sensor for exhaled CO analysis, it achieved a 23-fold sensitivity enhancement.
• Effectively tracked system states and eliminated parameter variations in dynamic systems.

Conclusions:

• The DO-AKF algorithm significantly enhances sensitivity in laser spectroscopy for trace gas detection.
• This method offers a robust solution for real-time monitoring in diverse fields.
• Potential applications include environmental pollutant monitoring, industrial process control, and breath gas diagnosis.