Non-Gaussian Systems Control Performance Assessment Based on Rational Entropy.

Area of Science:

• Control engineering
• Statistical process control
• Information theory

Background:

• Control loop Performance Assessment (CPA) is crucial for system operations.
• Stochastic statistical CPA indices, like minimum variance controller (MVC)-based indices, are widely used.
• Existing methods may not fully capture the complexities of non-Gaussian systems.

Purpose of the Study:

• To propose a novel Minimum Entropy Controller (MEC)-based CPA method for linear non-Gaussian systems.
• To enhance the description of random variable characteristics and uncertainty.
• To improve the estimation of performance benchmarks.

Main Methods:

• Utilizing probability density function (PDF) to describe system characteristics.
• Employing rational entropy (RE) to quantify the uncertainty of random variables.
• Implementing an improved Estimation of Distribution Algorithm (EDA) for parameter and noise PDF estimation.

Main Results:

• The proposed MEC-based CPA method effectively assesses performance in linear non-Gaussian systems.
• The use of PDF and RE provides a robust way to handle system uncertainties.
• The improved EDA algorithm enhances the accuracy of system parameter and noise PDF estimation.

Conclusions:

• The novel MEC-based CPA method offers a valuable advancement for non-Gaussian system analysis.
• This approach provides a more comprehensive understanding of system performance and uncertainty.
• The study demonstrates the practical applicability through ARMAX system case studies.