Ultrasound-based identification of damage in wind turbine blades using novelty detection.

Among the renewable energy sources, wind power generation presents competitive costs and high installation potential in many countries. Ensuring the integrity of the generation equipment plays an important role for reliable energy production. Therefore, nondestructive test procedures are required, especially for turbine blades, which are subject to severe operational conditions due to phenomena such as lightning strikes, mechanical stress, humidity and corrosion. Nondestructive ultrasonic test techniques are commonly applied in their predictive maintenance. This work proposes the use of novelty detection methods combined with nondestructive ultrasound testing to identify structural problems in wind turbine blades. Ultrasound signals are preprocessed using both, wavelet denoising and principal component analysis. Novelty detection deals with the one-class classification problem, when only the normal condition signatures are required for the classification system design. For the nondestructive test of turbine blades, this is an interesting paradigm because it is not always possible to obtain test samples from all of the existing flaw conditions. Our experimental results indicate the efficiency of the proposed method.

Segmented analysis of time-of-flight diffraction ultrasound for flaw detection in welded steel plates using extreme learning machines.

This work investigates the application of extreme learning machine, a fast training neural network model, for an ultrasound nondestructive evaluation decision support system. A novel segmented analysis of time-of-flight diffraction ultrasound signals is proposed in order to produce high flaw detection efficiency and low computational requirements, making it possible to be used in embedded applications. The frequency contents of TOFD signals temporal segments, estimated using the discrete Fourier transform, were used to feed the classification system. The test objects consisted of a set of SAE 1020 welded carbon steel plates, in which occur four types of defects. The obtained experimental results indicate that the proposed method is able to combine high accuracy, fast training and full exploration of the TOFD signal information.

Carburization level identification in industrial HP pipes using ultrasonic evaluation and machine learning.

Ultrasound nondestructive testing is commonly applied in industry to guarantee structural integrity. HP steel pyrolysis furnaces are used in petrochemical industry for lightweight hydrocarbon production. HP steel chromium content may be reduced in high-temperatures due to carbon diffusion. This characterizes the carburization phenomenon, which modifies magnetic properties, reduces mechanical resistance and may lead to structural rupture. For safe operation it is required to frequently determine carburizing level in pyrolysis furnace pipes. This is traditionally performed manually using magnetic evaluation. This work proposes a novel procedure for carburizing level estimation using ultrasonic evaluation associated to signal processing and machine learning techniques. Experimental data from pulse-echo ultrasonic tests performed in HP steel pipes are used. Discrete Fourier transform was applied for feature extraction and different classification systems (neural networks, k-nearest neighbors and decision trees) are applied and compared in terms of carburizing level identification efficiency.

Decision support system for ultrasound inspection of fiber metal laminates using statistical signal processing and neural networks.

The growth of the aerospace industry has motivated the development of alternative materials. The fiber-metal laminate composites (FML) may replace the monolithic aluminum alloys in aircrafts structure as they present some advantages, such as higher stiffness, lower density and longer lifetime. However, a great variety of deformation modes can lead to failures in these composites and the degradation mechanisms are hard to detect in early stages through regular ultrasonic inspection. This paper aims at the automatic detection of defects (such as fiber fracture and delamination) in fiber-metal laminates composites through ultrasonic testing in the immersion pulse-echo configuration. For this, a neural network based decision support system was designed. The preprocessing stage (feature extraction) comprises Fourier transform and statistical signal processing techniques (Principal Component Analysis and Independent Component Analysis) aiming at extracting discriminant information and reduce redundancy in the set of features. Through the proposed system, classification efficiencies of ~99% were achieved and the misclassification of signatures corresponding to defects was almost eliminated.