Resolving forward and inverse problems of rarefied gas heat transfer in an infrared detector cryochamber using physics-informed neural networks

This study presents a comprehensive analysis of rarefied heat transfer in cryogenic chambers with implications for infrared detector applications, using physics-informed neural networks (PINNs). Steady-state and transient heat transfer are analyzed to evaluate the steady cooling load and cooldown time as performance metrics in cryogenic chambers. We first developed a PINN-based framework to solve forward problems in rarefied gas heat transfer, presenting results by varying material properties and operating conditions such as thermal conductivity, emissivity, specific heat, rarefied gas pressure, and environmental temperature. The proposed framework is then extended to solve inverse problems, determining thermal conductivity and rarefied gas pressure based on operational requirements for steady cooling load and cooldown time in cryogenic chambers. Systematic analysis confirms that the proposed PINN-based framework successfully resolves both forward and inverse problems in rarefied gas heat transfer. We expect that the framework can be employed for the design of reliable cryogenic chambers and performance predictions under various environmental conditions.