Abstract
Establishing an accurate coarse tracking model is an important prerequisite for designing composite axis control systems and selecting devices for precision tracking. This task is complicated by measured operating data containing non-differentiable nonlinear links and unknown interference, which hinder standard system identification. We propose a method combining theoretical analysis and simulation modeling based on measured data. The model structure is determined through analysis, while specific parameters, nonlinearities, and interference are refined via simulation. The resulting model matches measured data well, providing a highly accurate nonlinear representation of the system.
Introduction
In high-precision laser tracking and pointing, engineers utilize a "composite axis" system. This architecture employs a nested setup: a large, heavy "coarse" turntable handles primary movements, while a smaller, agile Fast Steering Mirror (FSM) manages high-frequency "precision" corrections. To design and optimize the FSM, a high-fidelity mathematical model of the coarse system is required. However, real-world data is often "messy," containing non-differentiable nonlinearities like mechanical backlash and unknown signal interference. Standard linear system identification techniques struggle in these environments as nonlinearities pollute the linear model, leading to significant inaccuracies.
Methodology
The researchers developed a progressive, three-step method to build a high-fidelity model using limited measured operating data:
- Data Analysis: By plotting measured input, output, and error data, the team identified characteristic "jumps" and "jitter." These patterns indicated that the system was not purely linear but included mechanical backlash and internal interference.
- Structural Selection: While starting with a standard second-order system, the team noted that the actual system bandwidth was higher than expected. This necessitated the addition of a speed feedforward link (represented as a zero in the transfer function) to accurately reflect the response speed.
- Iterative Simulation Modeling: Using MATLAB/Simulink, the authors constructed a block diagram and iteratively adjusted parameters—such as gain ($k$), backlash magnitude, and interference types—until the simulation's error matched the real-world measured error sequence.
Key Insights and Results
The analysis revealed that the large, regular jumps in the error signal were the "fingerprint" of mechanical backlash, measured at 41 arcseconds. Furthermore, the high-frequency jitter indicated that interference was entering the system within the closed-loop rather than through the input signal.
The final nonlinear closed-loop transfer function was determined to be: $$G(s) = \frac{57.8455^2(1 + 0.0242s)}{s^2 + 2 \times 0.69 \times 57.8455s + 57.8455^2}$$
Key findings include:
- Feedforward Necessity: The discrepancy between physical bandwidth (7-10Hz) and equivalent data bandwidth (~220Hz) confirmed the presence of a speed feedforward link.
- Interference Characterization: The interference was successfully modeled as a combination of a non-unit random number and a 2Hz sine wave.
- Model Accuracy: Validation showed that the simulated error and measured error were highly matched, providing a reliable foundation for the subsequent design of the precision tracking system.
Conclusion
This study demonstrates that a hybrid approach combining analytical structural determination with iterative simulation can effectively model complex, nonlinear tracking systems. While the current parameter tuning process is manual, it successfully identifies critical hardware characteristics like backlash and internal noise that linear methods miss. Future work may involve automating this optimization process to further streamline the development of composite axis control systems.
Summary updated