An Improved Black-Winged Kite Algorithm for Global Optimization and Fault Detection.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Optimization Algorithms

Background:

• Accurate industrial fault detection is vital for production safety and efficiency.
• Feature selection (FS) methods are crucial for enhancing fault detection performance.
• The Black-winged Kite Algorithm (BKA) faces challenges like premature convergence in high-dimensional spaces.

Purpose of the Study:

• To propose an improved Black-winged Kite Algorithm (IBKA) to overcome the limitations of the original BKA.
• To enhance the performance of feature selection for industrial fault detection.
• To improve the convergence speed and robustness of optimization algorithms.

Main Methods:

• Introduction of a Stagnation-Triggered Diversification Mechanism to prevent local optima.
• Implementation of an Adaptive Weak Guidance Mechanism for improved convergence efficiency.
• Evaluation on 26 benchmark functions and real-world fault detection datasets using the binary version (bIBKA).

Main Results:

• IBKA demonstrated superior performance in solution quality, convergence speed, and robustness compared to BKA and other meta-heuristics.
• The binary IBKA (bIBKA) showed significant improvements in detection accuracy and reliability in fault detection applications.
• The enhancement mechanisms effectively balanced exploration and exploitation while maintaining simplicity and efficiency.

Conclusions:

• The proposed IBKA effectively addresses the limitations of the original BKA in high-dimensional feature selection.
• IBKA offers a robust and efficient solution for industrial fault detection.
• The novel enhancement mechanisms contribute to improved optimization performance and practical applicability.