Hydrogen induced cracking in energy pipelines and its monitoring with acoustic emission

Abstract

This research investigated the development of Hydrogen Introduced Cracks (HIC) on steel plates monitored with Acoustic Emission (AE) technology in detail. The work focuses on identifying and clustering HIC signals from mixed signals with different experimental setups, as well as describing HIC development in the light of signals’ characteristics. Electrochemical hydrogen charging (ECHC) method was applied to generate HIC on A516 steel plates with mixed solution of 0.5mol/L H2SO4 and 0.5g/L NaAsO2. Four defect mechanisms were found to generate elastic waves during a test, which were H2 evolution, HIC, crevice corrosion and uniform corrosion. An experimental procedure was designed for pattern recognition of these mixed signals. Short-time tests with or without current were carried out using only 0.5mol/L H2SO4 solution for identifying non-HIC signals separately. The energy proportions in the frequency range of 0-100kHz (PE1) and 100-200kHz (PE2) in the energy spectrum as well as the parameters in the time domain of the Duration, the Energy and the Counts are the main characteristics used for pattern recognition. These characters were the basic parameters for signals identification, irrespective of the profile of the plate, the value of applied current, the locations of the sensors. Due to the large amount of data, manual classification would be extensively time-consuming. In this study, a two-step Gaussian Mixed Model (GMM) clustering method was proposed for automatically clustering the mixed signals, in which the parameters of PE1 and PE2 obtained from the frequency domain were used for identifying signals from corrosions, and the Duration, the Counts and the Energy were then used to distinguish signals between H2 evolution and HIC. The effectiveness of this method was verified by experiments with different setup parameters. For the two types of sensors (Nano30 and VS150-RSC) used in this study, it was found that more components of the signals acquired by VS150-RSC were concentrated at 150kHz compared with those by Nano30 sensors. Therefore, the accurate rate of the proposed automatic clustering method was lower due to the scattering frequency distribution among the critical signals. However, sensor VS150-RSC is more suitable for HIC monitoring in situations of a low signal-to-noise ratio or over a long distance between the event and the sensor. For the different thickness of the specimen (5mm, 10mm and 20mm), it was found that the HIC signals acquired by VS150-RSC had components between 160-190kHz when the thicknesses of specimens were 10mm or 20mm. For the test with a complex structural profile including a hole and a seam of the same size, it was found that the seam heavily influenced the frequency distribution of HIC signals while a hole only influenced the characteristics of signals in the time domain. This research also investigated the source localisation of HIC events under different experimental setups, including the specifications of specimens and the type of sensors. The Simplex method was used to calculate the location based on the parallelogram sensor array. The onset time of each signal was determined by the Akaike Information Criterion (AIC) method to improve the accuracy. This method located HIC events well on simple plates with different setups but not on complex structures, such as a structure with holes, seams, welds, flanges and others. The delta-T mapping method was then investigated to localise HIC events on the complex plate with a hole and a seam. A FE model was also built by Abaqus to simulate the delta-T maps for each pair of sensors under this setup. Compared to the location results calculated by the Simplex method, the accuracy of localisation obtained by overlapping the simulated delta-T maps was greatly improved.