A general surrogate-based optimization algorithm for problems with computationally-expensive constraints with...

Area of Science:

• Control Engineering
• Optimization Algorithms
• Robust Control Systems

Background:

• Designing robust passive fault-tolerant controllers often uses μ-synthesis with manually set weight functions.
• Existing automatic tuning methods for these weights face challenges like local minima and computationally intensive constraints.

Purpose of the Study:

• To propose a novel surrogate-based global optimization algorithm for designing robust passive fault-tolerant controllers.
• To address limitations of existing automatic tuning methods, specifically nested optimization problems.

Main Methods:

• Development and demonstration of a novel surrogate-based global optimization algorithm.
• Comparison of the proposed algorithm against local optimization with multi-start and genetic algorithms.
• Application to automatic tuning of performance weights in μ-synthesis for robustness to multiplicative faults.

Main Results:

• The proposed algorithm demonstrates superior effectiveness and efficiency compared to benchmark methods.
• Numerical results assess performance concerning plant order and uncertainty levels.
• Designed controllers exhibit robust properties on a simulated mechatronic system under multiplicative faults.

Conclusions:

• The novel surrogate-based global optimization algorithm effectively overcomes limitations in designing robust passive fault-tolerant controllers.
• The method offers a computationally efficient and superior alternative for tuning control design parameters.
• The approach is validated for practical engineering applications, particularly in handling parametric uncertainties.