MFL based advanced condition assessment for aged cast iron pipes
- Publication Type:
- Thesis
- Issue Date:
- 2016
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Demanding essential industries such as water, petrochemical and energy use billions of dollars worth of metallic pipe infrastructure. Sydney Water Corporation alone has buried systems valued at over $15 billion and this is typical for large utilities. Catastrophic failures of these critical high pressure water pipes has severe impact on the general public, including disturbance to water supply, road traffic, compromised safety and surrounding infrastructure costing billions to the economy.
This raises the specific demand of accurate and reliable in-service inspection and maintenance of supply lines that do not disturb the service. A vibrant collaboration of researchers from Monash University, University of Technology Sydney and University of Newcastle along with ten water utilities and research institutes from around the globe are dedicated to research on ’The Advanced Condition Assessment and Pipe Failure Prediction Project’ through a collaborative research agreement and committed overall funding of $16 million.
Condition assessment of Cast Iron (CI) water mains is a major part of the project where non-destructive evaluation (NDE) technologies are considered. Magnetic Flux Leakage (MFL) technology has been a popular choice in the industry for NDE for decades, and hence it is a focus technology of this project. Obtaining accurate and reliable interpretations using MFL signals is a significant challenge due to the ill-posed nature of the inverse problem. Therefore the main contributions of this thesis stem from the use of different interpretation techniques to solve the MFL inverse problem.
In this thesis, initially a realistic analytical model was developed to capture the behaviour of MFL signals. Analytical model has restrictive assumptions and could not be utilised to solve the MFL inverse modelling. Therefore a data analytic approach using Gaussian Processes (GP) is proposed to capture highly complex MFL inverse model. The caveat in this method is that it is highly dependent on the quality of training data. To generate the required exhaustive data, a more generalised FEA based simulation model was developed and validated. The simulation model enabled the generation of exhaustive data to train a GP model and the model predictions were validated using the simulation data. This raised the need for experimental validation of the models and therefore the simulation model was experimentally prototyped. With successful experimental validations using the prototyped MFL lab set-up, a software module was produced to interpret real life industrial MFL data. This software produces 2.5D cylindrical approximated defect profile along with the uncertainty for each prediction.
Stress and failure analysis literature shows that ellipsoidal defects are the most vulnerable stress failure sources on gray cast iron pipes. In order to approximate ellipsoidal defects on aged CI pipes, a global optimisation procedure has been proposed. Using an analytical model the global optimisation algorithm produces the optimum ellipsoidal parameters which minimises the mismatch between the measurement and model output. The proposed algorithm produced high quality ellipsoidal parameters which can readily be used in the stress analysis.
MFL interpretation results were not only used for stress analysis, but also for multi-sensor fusion within the project. In this way, complementary information from different sensing modalities can be fused to enhance the overall prediction quality. Approximated cylindrical defects and ellipsoidal defects provided sparse information but this was not sufficient for multi-sensor fusion. A dense representation of the entire scan area is highly desirable. Further investigations of the MFL signals indicated that they contain information to interpret continuous thickness profiles. A global optimiser based on a FEA model has been iteratively used to estimate the continuous thickness profile. The results indicate that the proposed framework can accurately interpret dense 2.5D thickness maps.
This thesis addressed the MFL data interpretation problem in 3 specific scenarios; cylindrical defects approximation, ellipsoidal defects approximation and continuous remaining wall thickness profiling. Simulations, lab experiments and field trials were carried out for each scenario and the results were used to validate the effectiveness of the algorithms.
Please use this identifier to cite or link to this item: