Multi-Parameter Collaborative Optimization of a Coaxial Pulse Tube Cryocooler Integrating Acoustic–Mechanical–Electrical Coupling and Physical Boundary Constraints

Complex parameter interactions and acoustic–mechanical–electrical (AcME) impedance mismatches restrict coaxial pulse tube cryocooler (CPTC) performance, whereas previous studies typically optimized isolated cold fingers without compressor dynamic coupling. This study proposes a physics-prior-driven collaborative optimization framework for an integrated AcME-coupled CPTC system. Random Forest screening first compresses the 16-dimensional space to 10 core variables. Across 250 initial samples, the Kriging surrogate achieves baseline determination coefficients of R2=0.9812 for cooling capacity and R2=0.9663 for exergy efficiency. Although initial active learning expands the dataset to 350 samples, uncalibrated frontier exploration still encounters boundary extrapolation errors, inducing solver divergence and an initial 21.82% exergy discrepancy. To prevent search oscillations across non-convergent domains, a sequential active learning loop dynamically couples Constrained Expected Improvement infilling with Pareto feedback, while a Support Vector Machine enforces a numerical solvability boundary (Psafe≥0.85) alongside a power factor threshold (PF≥0.95). Computations deploy an 11-node coarse grid for search iterations and a 50-node fine grid for thermodynamic evaluation. Under fine-grid validation at 80 K, net cooling capacity reaches 14.41 W, and exergy efficiency increases from 14.65% to 17.65%, establishing a verifiable methodology for coupled thermodynamic systems.