Effective intelligent condition monitoring, as an effective technique to enhance the reliability of wind turbines and implement cost-effective maintenance, has been the object of extensive research and development to improve defect detection from supervisory control and data acquisition (SCADA) data, relying on perspective signal processing and statistical algorithms. The development of sophisticated machine learning now allows improvements in defect detection from historic data. This paper proposes a novel condition monitoring method for wind turbines based on Long Short-Term Memory (LSTM) algorithms. LSTM algorithms have the capability of capturing long-term dependencies hidden within a sequence of measurements, which can be exploited to increase the prediction accuracy. LSTM algorithms are therefore suitable for application in many diverse fields. The residual signal obtained by comparing the predicted values from a prediction model and the actual measurements from SCADA data can be used for condition monitoring. The effectiveness of the proposed method is validated in the case study. The proposed method can increase the economic benefits and reliability of wind farms.

The accumulation of fouling within a structure is a well-known and costly problem across many industries. The build-up is dependent on the environmental conditions surrounding the fouled structure. Many attempts have been made to detect fouling accumulation in critical engineering structures and to optimize the application of power ultrasonic fouling removal procedures, i.e., flow monitoring, ultrasonic guided waves and thermal imaging. In recent years, the use of ultrasonic guided waves has been identified as a promising technology to detect fouling deposition/growth. This technology also has the capability to assess structural health; an added value to the industry. The use of ultrasonic guided waves for structural health monitoring is established but fouling detection using ultrasonic guided waves is still in its infancy. The present study focuses on the characterization of fouling detection using ultrasonic guided waves. A 6.2-m long 6-inch schedule 40 carbon steel pipe has been used to study the effect of (Calcite) fouling on ultrasonic guided wave propagation within the structure. Parameters considered include frequency selection, number of cycles and dispersion at incremental fouling thickness. According to the studied conditions, a 0.5 dB/m drop in signal amplitude occurs for a fouling deposition of 1 mm. The findings demonstrate the potential to detect fouling build-up in lengthy pipes and to quantify its thickness by the reduction in amplitude found from further numerical investigation. This variable can be exploited to optimize the power ultrasonic fouling removal procedure.

There is an increasing interest in using ultrasonic guided waves to assess the structural degradation of above-ground storage tank floors. This is a non-invasive and economically viable means of assessing structural degradation. Above-ground storage tank floors are ageing assets which need to be inspected periodically to avoid structural failure. At present, normal-stress type transducers are bonded to the tank annular chime to generate a force field in the thickness direction of the floor and excite fundamental symmetric and asymmetric Lamb modes. However, the majority of above-ground storage tanks in use have no annular chime due to a simplified design and/or have a degraded chime due to corrosion. This means that transducers cannot be mounted on the chime to assess structural health according to the present technology, and the market share of structural health monitoring of above-ground storage tank floors using ultrasonic guided wave is thus limited. Therefore, the present study investigates the potential of using the tank wall to bond the transducer instead of the tank annular chime. Both normal and shear type transducers were investigated numerically, and results were validated using a 4.1 m diameter above-ground storage tank. The study results show shear mode type transducers bonded to the tank wall can be used to assess the structural health of the above-ground tank floors using an ultrasonic guided wave. It is also shown that for the cases studied there is a 7.4 dB signal-to-noise ratio improvement at 45 kHz for the guided wave excitation on the tank wall using shear mode transducers.

Ultrasonic guided waves are widely used to inspect and monitor the structural integrity of plates and plate-like structures, such as ship hulls and large storage-tank floors. Recently, ultrasonic guided waves have also been used to remove ice and fouling from ship hulls, wind-turbine blades and aeroplane wings. In these applications, the strength of the sound source must be high for scanning a large area, or to break the bond between ice, fouling and plate substrate. More than one transducer may be used to achieve maximum sound power output. However, multiple sources can interact with each other, and form a sound field in the structure with local constructive and destructive regions. Destructive regions are weak regions and shall be avoided. When multiple transducers are used it is important that they are arranged in a particular way so that the desired wave modes can be excited to cover the whole structure. The objective of this paper is to provide a theoretical basis for generating particular wave mode patterns in finite-width rectangular plates whose length is assumed to be infinitely long with respect to its width and thickness. The wave modes have displacements in both width and thickness directions, and are thus different from the classical Lamb-type wave modes. A two-dimensional semi-analytical finite element (SAFE) method was used to study dispersion characteristics and mode shapes in the plate up to ultrasonic frequencies. The modal analysis provided information on the generation of modes suitable for a particular application. The number of point sources and direction of loading for the excitation of a few representative modes was investigated. Based on the SAFE analysis, a standard finite element modelling package, Abaqus, was used to excite the designed modes in a three-dimensional plate. The generated wave patterns in Abaqus were then compared with mode shapes predicted in the SAFE model. Good agreement was observed between the intended modes calculated in SAFE and the actual, excited modes in Abaqus.

No abstract