Beyond the Drunken Walk : an Enhanced Flower Pollination Algorithm for Engineering Applications.

Koay, Hong Jun (2025) Beyond the Drunken Walk : an Enhanced Flower Pollination Algorithm for Engineering Applications. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Koay Hong Jun - FULL TEXT.pdf Restricted to Registered users only Download (3MB)

Abstract

This study investigates the application of Drunken Flower Pollination Algorithm (DFPA) which is the improved model of Flower Pollination Algorithm (FPA). It is a nature-inspired algorithm based on the pollination process of flowering plants and behavior of inebriated people to solve complex engineering problems. DFPA's capability to balance global exploration and local exploitation through the use of global pollinators (biotic pollination) and local pollinators (abiotic pollination) makes it particularly effective in finding optimal solutions in multi-dimensional search spaces. This research applies DFPA to critical engineering challenges, with a specific focus on the manufacturing sector, particularly in the detection of irregularities in centrifugal pumps using vibration and sound data. The algorithm’s performance is evaluated against three conventional feature-based classification methods. A hybrid approach integrating DFPA with Principal Component Analysis (PCA) achieves an average accuracy of approximately 94.45% across three engineering domains. Furthermore, experimental results indicate that sound-based data yield higher detection accuracy than vibration-based data in a noiseless environment when analyzed using the DFPA + PCA method. These findings highlight the potential of DFPA as a robust and effective tool for improving the precision and reliability of engineering diagnostics and optimization tasks.