Enhanced structural health monitoring of wind turbine blades using linguistic feature based crack detection and cuckoo optimized modular neural networks

Abstract

Wind power has emerged as one of the most widely used renewable energy sources because of its capacity to generate inexpensive, clean electricity at a reduced cost. To harvest a higher production of power, a larger wind turbine is needed. Because of its large construction, regular drawbacks are inevitable. The majority of the materials used to make the wind turbine blades are fiberglass. These blades are vulnerable to surface cracks due to wind breeze, poor environmental conditions, gravity loads, lightning strikes, and irregular wind loads. System integrity may suffer drastically from minimal structural damage. It is crucial to automate the inspection procedure and lower the level of uncertainty in routine blade health inspections. Hence, a structural health monitoring system that is economical, dependable, and predictive needs to be implemented. The primary goal of structural health monitoring in wind farm operations is to discover wind turbine blade faults early on by identifying cracks in the blades. In this work, a conventional linguistic feature-based crack recognition method is created for wind turbine blade structural health monitoring. The Modified Homomorphic Filtering (MHF), Local Binary Pattern (LBP), and Extreme Learning Machine (ELM) classifier has been proposed for the identification of linear cracks in wind turbines. Also, this thesis proposes a well established non-conventional heuristic Cuckoo Optimized Modular Neural Networks (COMNN) algorithm which relies upon the breedingetiquette of cuckoo species to locate the early cracks in wind turbine blades. newline