Optimal Sensor Placement via a POD-QR Framework for High-Fidelity 3D Temperature Field Reconstruction in Large-Scale...

Area of Science:

• Cryogenics and Thermal Engineering
• Data Science and Machine Learning
• Biobanking and Cryopreservation Technology

Background:

• Accurate temperature monitoring in ultra-low temperature (ULT) chest freezers is vital for biospecimen integrity.
• Dense sensor arrays for precise temperature mapping are often impractical due to space and cost limitations in large-scale freezers.
• Existing sensor deployment methods may not be optimal for large-volume cryogenic storage.

Purpose of the Study:

• To develop a systematic, data-driven framework for optimal sensor placement in large-scale ULT chest freezers.
• To enable high-fidelity cryogenic temperature field reconstruction using a sparse sensor network.
• To provide a cost-effective and space-efficient solution for thermal monitoring in biobanking.

Main Methods:

• High-resolution reference temperature fields were generated using universal kriging interpolation and validated with leave-one-out cross-validation (LOOCV).
• Principal component analysis (PCA) was employed to extract a proper orthogonal decomposition (POD) basis from training data.
• Optimal sensor locations were identified using QR-column pivoting on the POD basis, significantly reducing the number of required sensors.

Main Results:

• The kriging interpolation achieved a mean absolute error (MAE) ≤0.67 °C and R²>0.92.
• A sparse network of just 3 sensors was determined to be optimal, representing a 94% reduction from a 48-sensor setup.
• The optimized sparse sensor network enabled highly accurate temperature field reconstruction with grid-level MAE ≤0.079 °C and point-level MAE ≤0.502 °C.

Conclusions:

• The proposed data-driven framework effectively determines optimal sparse sensor subsets for large-scale ULT chest freezers.
• This approach enables reliable 3D cryogenic temperature field reconstruction and efficient thermal monitoring.
• The method offers a scalable, scientifically rigorous alternative to empirical standards, balancing accuracy, space, and cost for cryogenic biobanking.