RBF Neural Network‐Compensated Adaptive Sliding Mode Control for Elevator Vibration Suppression

To address lateral vibration challenges in elevator systems during operation, a high‐fidelity coupled dynamic model integrating the elevator car‐guide rail structure and voice coil motor (VCM) actuator is developed. Building on this framework, a radial basis function (RBF) neural network‐compensated variable‐speed sliding mode control (VS‐SMC) strategy is proposed, where the RBF network enables real‐time identification and compensation of critical nonlinear states and disturbances. A dsat switching function with adaptive zero‐point slope is designed as the variable‐rate reaching law, The synergistic operation of RBF network’s real‐time nonlinear compensation and dsat function’s adaptive slope modulation enables high‐precision suppression of elevator vibrations, with closed‐loop stability rigorously proven via Lyapunov theory. Under VCM operational constraints, comparative simulations in MATLAB/Simulink evaluate four control schemes: passive control, PID, tanh‐RBF SMC, and the proposed dsat‐RBF SMC. Results demonstrate superior vibration suppression with 97%, 41%, and 78% displacement reduction under sinusoidal excitation, and 92%, 76%, and 68% reduction under Gaussian white noise, relative to passive, PID, tanh‐RBF methods. The proposed method achieves minimum displacement responses across all test scenarios, validating its robustness and precision in real‐world elevator dynamics. This study establishes a neural‐enhanced sliding mode paradigm for high‐performance vibration control, offering critical insights for safety‐critical vertical transport systems.