Actuator‐Constraint Handling Versus Frequency Retuning in Active Seismic Control: A Two‐Height Yielding Steel‐Frame Benchmark

Actuator saturation forces two controller‐design choices: how to impose the force limit and how to represent post‐yield softening. We isolate the two choices on a reproducible OpenSeesPy reimplementation of the 3‐story and 9‐story Ohtori SAC steel‐frame benchmark structures, checked against published modal frequencies and selected uncontrolled response peaks. A matched finite‐horizon two‐by‐two factorial crosses elastic versus softened design models with post‐hoc clipping versus in‐QP actuator bounds. Across 44 3‐story evaluations using 22 FEMA P695 far‐field records at two intensity tiers, constraint handling accounts for 82% of the descriptive variance among four pooled cell‐mean costs, frequency tuning for 14%, and interaction for 4%; in the 9‐story 10‐case common‐finite‐score canonical subset the constraint‐handling share is 99.5%. Elastic in‐QP has the lowest mean under all five weight sets, yet under the primary score it loses to matched elastic clipping in 12 of 44 cases, with losses more frequent at larger uncontrolled drifts. Design‐frequency retuning stays small and inconsistent (elastic‐minus‐softened mean margin ); yield‐adaptive updating gives no consistent improvement, and a softening‐capable constitutive branch retains the principal ordering. In the transferred magnetorheological‐device comparison, powered operation reduces casewise peak drift by 19.9% on average but has a higher mean cost than zero‐current operation, and separate collapse‐limit analyses do not establish a consistent advantage for either allocation rule. For the primary score, the results prioritize actuator‐bound handling over the tested frequency retuning within the two planar frames, retaining clipping as a high‐demand comparator rather than a validated switching fallback.