Realtime optimization of multidimensional NMR spectroscopy on embedded sensing devices

Area of Science:

• Embedded Systems
• Sensor Technology
• Analytical Chemistry

Background:

• Embedded devices require autonomous sensor optimization due to limited computing resources.
• Current optimization methods are computationally intensive, hindering real-time application and leading to underutilization of deployed assets.

Purpose of the Study:

• To develop an efficient, on-the-fly sensor optimization routine for resource-constrained embedded devices.
• To improve measurement efficiency in applications like Nuclear Magnetic Resonance (NMR) relaxometry.

Main Methods:

• Implemented an Optimal Experimental Design (OED) routine partitioning sample properties into classes.
• Utilized supervised learning models to learn sensor signals for each class.
• Deployed trained models on constrained devices for real-time data processing and measurement optimization.

Main Results:

• Demonstrated a closed-loop optimization method for multidimensional NMR relaxometry on complex fluids.
• Achieved continuous classification and optimization of measurements with minimal computing load.
• Validated the method's suitability for remote sensing and IoT networks.

Conclusions:

• The proposed OED-based approach enables autonomous, real-time sensor optimization on embedded devices.
• This method significantly enhances measurement efficiency and asset utilization in field applications.
• The technique is well-suited for resource-limited environments, including remote sensing and IoT infrastructure.