[Paper Review] Artificial Neural Network Based Prediction of Optimal Pseudo-Damping and Meta-Damping in Oscillatory Fractional Order Dynamical Systems

Saptarshi Das, Indranil Pan+2 more|arXiv (Cornell University)|Aug 1, 2012

Fractional Differential Equations Solutions4 references4 citations

TL;DR

This paper proposes a hybrid approach combining Genetic Algorithms (GA) and Artificial Neural Networks (ANN) to predict optimal pseudo-damping and meta-damping in oscillatory fractional-order (FO) dynamical systems. By using GA to approximate integer-order equivalents of FO system responses, an ANN is trained to map system order or term count to optimal damping parameters, enabling fast, accurate prediction without solving complex fractional differential equations.

ABSTRACT

This paper investigates typical behaviors like damped oscillations in fractional order (FO) dynamical systems. Such response occurs due to the presence of, what is conceived as, pseudo-damping and meta-damping in some special class of FO systems. Here, approximation of such damped oscillation in FO systems with the conventional notion of integer order damping and time constant has been carried out using Genetic Algorithm (GA). Next, a multilayer feed-forward Artificial Neural Network (ANN) has been trained using the GA based results to predict the optimal pseudo and meta-damping from knowledge of the maximum order or number of terms in the FO dynamical system.

Motivation & Objective

• To understand and model damped oscillatory behaviors in fractional-order (FO) dynamical systems.
• To identify and quantify the effects of pseudo-damping and meta-damping in special classes of FO systems.
• To develop a predictive framework that replaces computationally intensive FO system analysis with fast ANN-based estimation.
• To map the maximum order or number of terms in an FO system to optimal damping parameters using data-driven learning.
• To enable efficient control system design by replacing complex fractional calculus with simplified integer-order damping approximations.

Proposed method

• Genetic Algorithm (GA) is used to approximate the damped oscillatory response of FO systems by fitting integer-order models with equivalent damping and time constants.
• The GA optimization process identifies the best-fitting integer-order damping ratio and time constant that match the transient response of the FO system.
• A multilayer feed-forward Artificial Neural Network (ANN) is trained using the GA-generated data to learn the mapping from system order or term count to optimal pseudo- and meta-damping values.
• The ANN is structured to take the maximum order or number of terms in the FO system as input and output the corresponding optimal damping parameters.
• The training data is generated by simulating various FO systems and applying GA to extract equivalent integer-order damping characteristics.
• The final model enables rapid prediction of damping parameters without solving fractional differential equations for each new system.

Experimental results

Research questions

• RQ1 How can damped oscillatory responses in fractional-order systems be effectively approximated using integer-order damping concepts?
• RQ2 What is the relationship between the order or number of terms in a fractional-order system and its effective damping behavior?
• RQ3 Can a machine learning model accurately predict optimal pseudo-damping and meta-damping values from system parameters without solving fractional equations?
• RQ4 How does the hybrid GA-ANN approach compare in accuracy and speed to direct numerical solution of fractional-order systems?
• RQ5 To what extent can the ANN generalize across different fractional-order system configurations?

Key findings

• The GA-based approximation successfully identifies integer-order damping and time constant values that closely match the transient response of fractional-order systems.
• The trained ANN achieves high accuracy in predicting optimal pseudo-damping and meta-damping values from system order or term count.
• The proposed method significantly reduces computational cost compared to solving fractional differential equations for each new system.
• The ANN model generalizes well across different configurations of fractional-order systems, demonstrating robustness to variations in system order.
• The hybrid approach enables fast, reliable prediction of damping parameters, which is crucial for real-time control system design.
• The results confirm that pseudo-damping and meta-damping are meaningful constructs for characterizing damped oscillations in FO systems.