I modeled an elastic substrate with axially stretched connectors in Abaqus. Initially, everything was defined in meter units (elastic modulus = 100 Pa, connector stiffness = 100 N/m). Later, I converted the entire model to millimeter units and updated the properties accordingly (elastic modulus = 100e-6 MPa, connector stiffness = 100e-3 N/mm). All geometric dimensions, connector lengths, and influence radius were scaled consistently.

Although I expected equivalent behavior, the connector in the millimeter-based model behaves much stiffer. I am using a distributing (rotational structural) coupling.

Is there anything beyond basic unit scaling (e.g., coupling formulation, connector definition, or hidden unit dependencies) that could explain this inconsistency?