Pattern recognition by wavelet transforms using macro fibre composites transducers

• Sensors based methodology with easy placement in complex surfaces.
• Use of a technique (wavelet methodology) that is able to work at different frequencies and for all types of signals.
• Sensitive pattern recognition for signals of low variability.
• Successful results for up to the 94% of the cases studied.

This paper presents a novel pattern recognition approach for a non-destructive test based on macro fibre composite transducers applied in pipes. A fault detection and diagnosis (FDD) method is employed to extract relevant information from ultrasound signals by wavelet decomposition technique. The wavelet transform is a powerful tool that reveals particular characteristics as trends or breakdown points. The FDD developed for the case study provides information about the temperatures on the surfaces of the pipe, leading to monitor faults associated with cracks, leaks or corrosion. This issue may not be noticeable when temperatures are not subject to sudden changes, but it can cause structural problems in the medium and long-term. Furthermore, the case study is completed by a statistical method based on the coefficient of determination. The main purpose will be to predict future behaviours in order to set alarm levels as a part of a structural health monitoring system.