Automatic data interpretation and enhanced localization for inline remote field eddy current tools
- Publication Type:
- Thesis
- Issue Date:
- 2018
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Most of the water pipelines laid in the 20th century, in Sydney, are made of ferromagnetic materials that corrode with time. Corrosion weakens the structure of the pipes and may eventually lead to catastrophic failures. Thus, regular inspection and maintenance of critical-pipes is needed which is both costly and challenging. Non-Destructive Evaluation (NDE) technologies, such as Remote-Field Eddy-Currents (RFEC), are a cost-effective option to assess the condition of pipes.
RFEC technology is based on the double-through wall phenomenon, which results in having different areas of the pipe's geometry being convoluted into the RFEC sensor measurements. Thus, the interpretation of the signal into thickness information is a challenging task. The technology is traditionally studied using Finite Element Analysis (FEA) for very simple geometries. Examples found in the literature tend to consider for instance perfect cylindrical pipes in the presence of square axisymmetric defects. In practice, these experiments do not translate well with the organic shapes generated by the corrosion, and these idealistic scenarios bypass the need for signal deconvolution. Furthermore, the behaviour of the tool in three-dimensional space is not well understood.
In this thesis, FEA simulations are performed on geometries obtained from real corroded pipes. Thus, the simulations are a reflection of a realistic RFEC inspection. Based on FEA, data-driven algorithms have been designed to solve the direct and inverse problems for homogeneous materials and to solve the signal deconvolution for nonhomogeneous materials (which requires an additional piecewise linear transformation to fully solve the inverse problem), in both, the two-dimensional axisymmetric scenario and the three-dimensional scenario. These algorithms have been tested on datasets obtained through simulations, as well as, field deployment of an inspection tool. Additionally, a localisation algorithm is proposed to align the RFEC data obtained from the field inspection with laser-scan measurements used as ground truth. Finally, a methodology for automating the data analysis for the extraction of defects present in RFEC data (in terms of localisation and 2D shape segmentation) has been developed and tested with real data. As a result, a framework is proposed to process raw RFEC data and ultimately extract the location and shape of defects which will, in turn, assist with pipe failure prevention.
Please use this identifier to cite or link to this item: