Advanced analytics in maintenance management based on ultrasonic guided waves

Concentrated Solar Power (CSP) and Wind Turbine (WT) energy are an alternative to the conventional energy sources. There is a significant rising in new and sophisticate CSP and WT farms due to global energy demands. They require improving the operational and maintainability in this industry. A proper predictive maintenance program is required to detect defects at an early stage, to reduce corrective maintenance costs and increase the reliability, availability, and safety. Recently, Structural Health Monitoring (SHM) systems and methods have been developed for Wind Turbine Blades (WTB) and CSP pipelines. SHM is a key technique for measuring and evaluating structural integrity in the new “smart-blade” concept, which involves the entire WTB life cycle, including design, operation, maintenance, repair, and recycling. Pattern recognition techniques play an important role in fault detection and diagnosis. Different damages or faults have been analysed in this research work for WTB and CSP: Delamination is a common problem in WTBs, leading stress areas that can generate the partial or complete rupture of the blade; dirt and mud on WTB reduce productivity and can generate stops and downtimes; ice on WTB increases the surface roughness and reduce the aerodynamic efficiency, generating an imbalance in the rotor that produces stress in WTB and drives train; the absorber tubes in normal working conditions must withstand high temperatures, that could cause faults, e.g. corrosion, in areas such as welding, or it can cause hot spots due to defects or corrosion. The thesis deals with a novel research on Advanced Analytical Methods in Maintenance Management based on Ultrasonic Guided Waves. The main objective of this research is to approach different techniques of Pattern Recognition using ultrasonic Guided Waves (GWs) applied to WTB and CSP. The aim of this research is to provide new solutions for locating, detecting and classifying damages in WTBs, e.g. delamination, ice, dirt and mud, or in CSP, e.g. notch, temperature analysis, etc. Pattern recognition methodology has been developed in two main blocks: Signal Processing techniques, e.g. Wavelet Daubechies Energy, Times Frequency Analysis and Energy Attenuation, and; Machine Learning by the analysis of Linear and Nonlinear models at different frequency levels