This document summarizes the process of modeling, designing controllers for, and implementing controllers on both the primary VCM actuator and secondary PZT micro-actuator in a dual-stage actuator system. For each actuator, system identification is performed to extract modal parameters from frequency response data. Notch filters are designed and controllers are implemented, tested in simulation, and tuned based on hardware implementation results. The goal is to meet stability and performance specifications for both individual actuators and eventual integration into a dual-stage system.
This document summarizes the process of modeling, designing controllers for, and implementing controllers on both the primary VCM actuator and secondary PZT micro-actuator in a dual-stage actuator system. For each actuator, system identification is performed to extract modal parameters from frequency response data. Notch filters are designed and controllers are implemented, tested in simulation, and tuned based on hardware implementation results. The goal is to meet stability and performance specifications for both individual actuators and eventual integration into a dual-stage system.
This document summarizes the process of modeling, designing controllers for, and implementing controllers on both the primary VCM actuator and secondary PZT micro-actuator in a dual-stage actuator system. For each actuator, system identification is performed to extract modal parameters from frequency response data. Notch filters are designed and controllers are implemented, tested in simulation, and tuned based on hardware implementation results. The goal is to meet stability and performance specifications for both individual actuators and eventual integration into a dual-stage system.
Single-Stage Actuators One Actuator operating at any one time Primary: VCM Actuator (PZT off) Secondary: PZT Micro-actuator (VCM off) Dual-Stage: Power up VCM + PZT
Project Scope - II For each actuator, perform: - Frequency Response Measurement - System Identification, G(s) - Notch Filter Design, F(s) - Controller Design, C(s) - Discretization, C(z)F(z) - Hardware Controller Implementation
Then design for Dual-Stage Actuator - Master-Slave Decoupling/PQ Method etc.
What has been done I VCM Frequency Response Measurement System Identification Notch Filter Design Controller Design Discretization Hardware Controller Implementation What has been done II PZT micro-actuator Frequency Response Measurement System Identification Notch Filter Design Controller Design Discretization
What has been done III Main focus: VCM - Explore - Simulate with different controllers - Compare simulation results with experiment - Analyze PZT simulations - Could be used by interested learners for hardware test
System Identification I VCM Experimental Frequency Response vs. Identified VCM Model
Model:
System Identification II Rigid body model - Repeated pole at around 1 kHz Resonant Modes - in-phase mode = 2 poles - out-of-phase mode = 2 zeroes Extraction of Modal Parameters
- natural frequency = frequency at resonant peak - small damping ratio - (half-power bandwidth) damping ratio ( a - b )/2 r
System Identification III Resonant Modes, i Resonant Frequency (kHz) Damping Ratio, i
1 4.43 0.0044 2 7.61 0.0087 3 8.92 0.0068 Modal Parameters of the identified VCM model Notch Compensation 4 Notch Filters Designed for the 4 resonant modes above 7 Notch Filters used initially Notch Filters brings about drastic phase changes Leads to implementation problem on DSP
Desired Frequency Domain for Controller Design - Closed-loop Stability: 6 dB GM, 40 deg PM - Disturbance Rejection: High gain at low frequency range - Low High frequency Gain: Reduce noise amplification PI-Lead Controller Design Integral Control with Lead Compensator Integrator: Infinite gain at low frequency Lead: Boosts PM at gain c/o frequency
Lag-Lead Controller Design I Lag: Increases low frequency gain Reduced steady state error Lag-Lead chosen over PI-Lead: unidentified problem with DSP during PI implementation Closed-loop Step Response I
Discretization and Simulink Source: unit square wave input 40 kHz sampling frequency Tustin approximation Hardware Implementation of the Lag-Lead Compensator I Undue vibration observed from the VCM plant Indication of Instability Large oscillations Problem Identification and Controller Redesign Mismatch between VCM bode plot and the plot of the identified model (slide 6) Gain lowered at low frequencies (slide 7) Pure gain added for adjustment (slide 8) Resonant modes are the same Lag-Lead Compensator Redesigned
Problem Identification contd. Problem Identification contd. Controller Redesign II (gain c/o frequency = 460 Hz) Closed-loop Step Response II Hardware Implementation of Lag- Lead Compensator II (gain c/o = 460 Hz) Small Bandwidth Slow Response Oscillations Bandwidth can be increased
Controller Design III (gain c/o frequency = 955 Hz) Closed Loop Step Response III Simulink Plots for Discrete-time I Source: Unit Square Wave Input 40 kHz Sampling frequency ZOH method Simulink Plots for Discrete-time II Source: Saw-tooth wave 40 kHz Sampling frequency Tustin method
Simulink Plots for Discrete-time III Source: Sine Wave 40 kHz Sampling Frequency Tustin method
Hardware Implementation of Lag- Lead Compensator III (gain c/o = 955 Hz) Faster response Increased oscillations Possible drawback in notch design Variation in VCM resonant modes Wider notch filter could be used
Control Design Results Open- loop transfer function Gain Margin (dB) Phase Margin (deg) Gain c/o Frequen cy (Hz) Oversho ot (%) Rise time (s)
Settling time (s) P vcm (s)N vcm (s)C vcm (s) (460 Hz) 6.18 52.9 460 22.8 0.000181 0.00174 P vcm (s)N vc m (s)C vcm (s) (955 Hz)
6.79 46.8 955 23.2 0.000186 0.00168 Desired Specificat ions > 6 > 30 1000 < 20 < 0.0002 < 0.002 PZT Micro-actuator: System Identification PZT control: Lag Compensator cascaded with 4 Notch Filters Closed-loop Step Response Control Design Results GM = 13.3 dB PM = 49.2 deg Overshoot = 12.8 % Settling time = 1.66 ms Discretization and Simulink Square wave input source 40 kHz Sampling frequency Tustin method