Dear @Zhizhao.Yin,
OpenFAST allows the aerodynamic discretization and reference axis to be different from the structural discretization and reference axis. In AeroDyn, BlCrvAC, BlSwpAC, and BlCrvAng should be based on the reference point for the aerodynamic pitching moment (which we call the aerodynamic center in AeroDyn), which is likely quite independent from any center (tension center, shear center, mass center) in the structural model. BlCrvAng represents the out-of-plane angle of the airfoil due to curvature, as explained in the online AeroDyn documentation on readthedocs: 4.2.3. Input Files — OpenFAST v4.1.1 documentation. It is also possible to derive BlCrvAng from BlCrvAC, as discussed in the following question on OpenFAST GitHub issues: Obtaining Blade geometry inputs BldSwpAC, BldCrvAC · Issue #451 · OpenFAST/openfast · GitHub.
In BeamDyn, I would choose a reference axis (kp_xr, kp_yr, kp_zr) that follows the natural geometry of the curved blade (if the blade is straight, I would choose the pitch axis as the reference axis; but for curved or swept blades, I would start the reference axis at the root at the pitch axis and choose a curve that naturally follows the curve and sweep) and define the sectional mass and stiffness matrices with respect to that. For an isotropic material, the shear and torsion stiffness terms are coupled independently from the axial and bending stiffness terms, but for an anisotropic composite material, the sectional stiffness matrix can be fully populated. The document (BeamDynInput_05_2020.pdf) attached in my post dated Jul 20, 2020 in the following forum topic explains that well: Simulations using BeamDyn.
Best regards,