Acceleration-Driven Observer-Based LuGre Modeling for Sensor-Reduced Identification of Magnetorheological Damper Dynamics

Modeling magnetorheological (MR) dampers for structural vibration mitigation remains challenging due to complex nonlinear hysteretic dynamics. Traditional parametric models rely on displacement and velocity inputs, often neglecting high-frequency inertial forces and requiring extensive sensing infrastructure. To address these limitations, an Acceleration-Dependent LuGre (ADLG) model explicitly utilizing piston acceleration is proposed. By augmenting the classical friction framework with an acceleration-driven internal state, a virtual mass, and current-modulated parameters, the ADLG formulation isolates hysteretic, rate-dependent, and inertial effects. This acceleration-centric approach simplifies instrumentation by eliminating velocity sensors and circumventing drift-prone numerical integration. Model parameters were identified via Particle Swarm Optimization (PSO) using experimental data from a LORD RD-1097-01 damper and validated in situ within a reduced-scale five-story structural frame. Benchmarking against five established models (Modified Bingham, Modified LuGre I and II, Modified Bouc-Wen, and Modified Algebraic) demonstrated highly competitive performance. During training, the ADLG model achieved an optimal R 2 = 0.9182 , a minimum RMSE of 9.11 N, and a 16.5% reduction in the 95th percentile of absolute error ( E 95 ) relative to the closest alternative. Validation across diverse deterministic multi-frequency excitations yielded localized steady-state R 2 values up to 0.9929. Furthermore, robust predictive capabilities were maintained under abrupt current steps, biased-sine modulations, and broadband seismic profiles. Explicit incorporation of acceleration dynamics establishes a robust, sensor-efficient framework for practical semi-active control implementations, positioning the ADLG as an excellent alternative to conventional formulations.