Decision Tree Pattern Recognition Model for Radio Frequency Interference Suppression in NQR Experiments

Area of Science:

• Nuclear Magnetic Resonance Spectroscopy
• Nuclear Quadrupole Resonance Spectroscopy
• Machine Learning Applications
• Signal Processing

Background:

• In-situ applications of unshielded nuclear quadrupole resonance (NQR) and nuclear magnetic resonance (NMR) are limited by radio frequency (RF) interference in industrial settings.
• Quality control and assurance applications require robust methods resistant to environmental electromagnetic noise.
• Burst-mode RF interference poses a significant challenge for sensitive spectroscopic techniques.

Purpose of the Study:

• To investigate the effectiveness of machine learning for identifying and mitigating radio frequency interference in NQR/NMR applications.
• To develop and validate an automated interference suppression algorithm for quality control and assurance.
• To assess the performance improvement in a contraband detection scenario using the proposed method.

Main Methods:

• A machine learning decision tree model was developed for automated identification of RF interference bursts.
• The developed algorithm was integrated to support automated interference suppression techniques.
• Receiver Operating Characteristic (ROC) analysis was employed to evaluate the performance of the new processing method against traditional approaches.

Main Results:

• The machine learning decision tree model demonstrated suitability for automated identification of interference bursts.
• Data processed with the new algorithm showed a significant improvement in accuracy compared to traditional methods.
• The area under the ROC curve increased from 0.580 to 0.906, indicating highly significant improvement in identifying data distorted by interference.

Conclusions:

• Machine learning, specifically decision tree models, offers a viable solution for automated detection and suppression of RF interference in NQR/NMR.
• The developed algorithm significantly enhances the reliability of NQR/NMR data analysis in noisy industrial environments.
• This approach shows promise for improving the performance of security applications, such as contraband detection, by overcoming RF interference limitations.