In the last two decades, nonlinear ultrasonic testing is getting more attention due to their sensitivity to microcracks among a variety of NDT techniques used in infrastructure. Vibro-Acoustic Modulation (VAM) technique is one of the practical methods, that does not need the expensive hardware components required for the conventional nonlinear methods. This method is capable of identifying damage growth using the correlation of the level of nonlinearity to the severity or density of the damage. To be able to determine the sensitivity of VAM technique in comparison with other conventional nondestructive testing methods, Acoustic Emission (AE) technique as a global method and Ultrasonic Testing (UT) and Eddy current Testing (ET) techniques as local methods are investigated in an identical testing condition for similar specimens. The comparison has been conducted by testing a typical steel material used in the steel bridges under cyclic tension load. All these methods have some features in common and some differences. A comprehensive comparison study of these techniques sheds light on their practicality for various applications. Unlike the AE technique that listens to the structure for the received signal of the released elastic energy from the defects, VAM introduces the signals to the specimen and monitors the signal that was modulated by the vibration to get information about the crack. VAM and AE have some similarities such as no need for positioning sensors on the cracks and capability of detecting the crack in the early stages. On the other hand, local techniques such as UT and ET are more accurate than the VAM technique in terms of localization but less sensitive in terms of how soon they detect cracks.
Boiler tubes in power plants develop defects including creep and thermal fatigue damage that can lead to fluid leakage over the operation period. Such leakage is the main cause of outages and power generation losses in thermal power plants. Therefore, early detection of leaks in boiler tubes is necessary to avoid more than 60% of boiler outages. To monitor and detect tube leaks in real-time, Acoustic Emission (AE) technique is widely used in power plants. A boiler tube leak could be detected using Average Signal Level (ASL) of the acquired AE signal using a network of sensors attached to the body of the boiler. Changes in ASL are proportional to the tube leakage; however, background signals generated by operating soot blowers bury the features which represent the tube leaks in the boiler and makes it nearly impossible to detect them automatically with established threshold methods. Soot blowers are used to remove the soot that is deposited on the tubes to maintain the efficacy and continuous operation of boilers. In this study, a bidirectional long short-term memory (LSTM) recurrent neural network (RNN) is developed to automatically detect tube leaks in power plant boilers. This detection method aims at identifying abnormal acoustic signals which differ from the reference/normal data that the system was trained with. The neural networks are trained on a sample boiler and the evaluation was done on the same boiler on the intervals with leak presence. Once the developed machine learning algorithm was tested with AE signals acquired from boiler tubes, the results show that this novel approach can detect anomalies in the signal levels as an indication of tube defects with an acceptable accuracy.
Ultrasonic Phased Array imaging is a key method for fast and reliable nondestructive testing of structures, especially when only one side of the part is accessible. Full matrix capturing (FMC) in combination with the total focusing method (TFM) provides a strong tool for ultrasonic imaging of structures with complex flaw patterns. However, still, operator needs to go through the generated images and manually check for the possible defects. One important task is to separate true and false indications, as some of them are noises or artifacts. Inspecting large structures with TFM Phased Array Imaging generates a huge amount of data which takes a significant time to go through them manually. In this work, we evaluate the possibility of using the neural network as an artificial intelligent toolbox to identify the defects. Using finite element method and an in-house developed TFM code, the phased array images are produced as the input to the neural network. The output of the neural network, target, is defined as the probability of defect existence. After generating TFM final images with different flaw patterns, the network was trained and evaluated based on the stochastic genetic algorithm learning method. This made the training feasible with limited provided data. Results indicate the great potential of machine learning for automatic or assisted defect recognition. The main challenge to pursuing a comprehensive and reliable machine learning toolbox, is to train the system with a satisfactory number of examples in different situations to ensure the final product is able to cover all possible conditions. It is concluded the proposed neural network model is capable of image pattern recognition with limited provided training data.