Wind Turbine Blade Monitoring via Deep Learning and Acoustic Aerodynamic Signals

Wind turbines have developed as an important and ecologically friendly source of power. However, owing to the extensive usage of innovative materials, ensuring the dependability of these devices has become crucial. The focus of research has shifted to the development of efficient and cost-effective monitoring methods for wind turbine blades, which are often the most expensive component of a wind turbine. In this study, we provide a novel monitoring system for wind turbine blades based on the medical auscultatory approach, which makes use of a deep convolutional neural network methodology. A balance is struck in the system's design between engineering dependability and economic efficiency. This study provides a lightweight framework for monitoring wind turbine blades that makes use of edge computing and signals given by the programmable logic controllers in the turbines. This system effectively gathers and filters crucial aerodynamic acoustic signals. Then, to reduce noise in wind turbine blade audio signals, we offer a set of audio enhancement approaches that include self-adaptive mask targeting, multi-scale feature extraction, and deep neural networks. We offer a unique method for compressing deep convolutional neural networks in our tertiary work, making them appropriate for peripheral computing devices with limited resources. In addition, we maximize the usage of audio-generated spectrograms for wind turbine blade issue diagnosis.