Predictive thermal model for indirect temperature measurement inside atomic cell of nuclear magnetic resonance gyroscope

This paper describes a methodology for predicting temperatures inside a micro-device at critical locations where thermal sensors cannot be placed. The device under study is a nuclear magnetic resonance gyroscope, an instrument that is highly sensitive to temperature variations. A physical thermal model of the device is developed and validated using experimental measurements from a prototype. A compact thermal model which approximates the behavior of the physical system is obtained through a sequential selection algorithm (with RMS error of approximately 0.15degC). The compact thermal model can be solved rapidly and is therefore suitable for real-time feedback control of the temperature inside the atomic cell.