A robust computational technique for model order reduction of two-time-scale discrete systems via genetic algorithms

Area of Science:

• Control Systems Engineering
• Computational Intelligence
• Systems Theory

Background:

• Multi-time-scale systems exhibit complex dynamics, often influenced by singular perturbation.
• Traditional model order reduction (MOR) methods may struggle to preserve essential system behaviors.
• Identifying and simplifying non-dominant dynamics is crucial for efficient system analysis.

Purpose of the Study:

• To develop a robust computational technique for model order reduction (MOR) of discrete multi-time-scale systems.
• To leverage genetic algorithms (GA) for MOR, ensuring preservation of dominant dynamics and minimizing steady-state error.
• To provide an effective approach for simplifying complex systems in engineering applications.

Main Methods:

• A novel MOR approach is proposed utilizing genetic algorithms (GA).
• The reduction process involves transforming the system's state matrix into an upper triangular form.
• GA optimizes the reduced-order model by maximizing a fitness function based on response deviation.

Main Results:

• The proposed GA-based MOR technique successfully generates reduced-order models.
• The method accurately maintains the dominant dynamics of the original multi-time-scale systems.
• Simulation results demonstrate significant advantages over existing MOR techniques, including minimized steady-state error.

Conclusions:

• The genetic algorithm-based computational intelligence approach offers a robust and advantageous method for MOR.
• This technique is effective for simplifying both single-input single-output (SISO) and multi-input multi-output (MIMO) discrete systems.
• The approach shows significant potential for applications requiring efficient analysis of multi-time-scale systems.