Gaussian Process-Based Multi-Fidelity Bayesian Optimization for Optimal Calibration Point Selection.

Area of Science:

• Metrology
• Environmental Engineering
• Optimization Techniques

Background:

• Temperature and humidity calibration chambers are crucial for aerospace and biomedicine.
• Traditional calibration methods suffer from limited range and low efficiency.
• Existing methods struggle to adapt to complex operational requirements.

Purpose of the Study:

• To develop an advanced framework for optimal selection of temperature and humidity calibration points.
• To address the limitations of traditional fixed calibration points in terms of adaptability, coverage, and efficiency.
• To enhance the reliability and accuracy of calibration chambers.

Main Methods:

• Development of a Gaussian Process-based Multi-Fidelity Bayesian Optimization (GP-MFBO) framework.
• Integration of a three-layer progressive multi-fidelity modeling system (analytical, CFD, experimental).
• Implementation of a systematic uncertainty quantification and an adaptive acquisition function.

Main Results:

• GP-MFBO achieved optimal calibration points with temperature uniformity of 0.149 and humidity uniformity of 2.38.
• Uniformity score improvements of up to 81.7% (temperature) and 76.3% (humidity) compared to other methods.
• Prediction confidence interval coverage reached 94.2%, surpassing comparative methods.

Conclusions:

• The GP-MFBO framework offers a robust solution for optimizing calibration point selection.
• This research provides a foundation for the scientific design of large-space temperature and humidity calibration systems.
• The proposed method enhances efficiency and accuracy in instrument testing and validation.