Nested Optimization of an Elevator and Its Gain-Scheduled LQG Controller

This paper studies the combined optimization of an elevator’s design (plant) and LQG controller for ride comfort. Elevator dynamics and primary vibration sources (drive motor torque ripple and guide rail irregularity) are modeled using an object-oriented language. The resulting model is nonlinear. Elevator vibrations are minimized with respect to both the design and the LQG controller. LQG gains are scheduled versus cab mass and height for robustness. Sequential plant/control optimization produces an optimal ride only when the torque ripple is the dominant disturbance. Otherwise, passive vibration reduction decreases the controller’s authority over the vibrations, hence coupling the plant and control optimization problems. Combined plant/controller optimization, using a nested strategy, mitigates this coupling and finds the correct optimal system design.