Autonomous defect classification for ultrasonic non-destructive testing in advanced manufacturing processes PhD Studentship
Ultrasonic Testing (UT) is a popular method in advanced manufacturing to evaluate the integrity of components without having an impact of their future usefulness (non-destructive testing). Commonly, components are inspected using manual encoded solutions (conventional UT or Phased Array Ultrasonic Testing (PAUT)) and the data is then analysed by a skilled operator. As such is dependent on the operator’s testing experiences and as this could potentially lead to errors, there is an academic research and industry need to develop a system capable of Autonomous Defect Recognition (ADR), potentially to be integrated alongside existing inspection processes and systems.
The overall objective of this PhD is to develop a computational framework for autonomous defect classification for ultrasonic testing. Research work will focus on Machine Learning (ML) methodologies (but not limited to ML alone), with potential investigation topics on feature extraction and selection methods in spatiotemporal signals, model-based supervised and unsupervised classification algorithms, as well as autonomous and semi-autonomous model-based decision making.
The overarching aim of the computational framework is to provide interpretation of detected signals within welds or bonds and to enable these to be classified according to a predetermined set of relevant acceptance criteria. This will necessitate the development of a system that seamlessly integrates to existing phased array equipment, and following an algorithmic process, displays the outcome of the autonomous classification results in simple linguistic terms for the process operators.
The computational framework to be developed will be assessed for its generalisation properties, based on a general selection of welds and bonds but also specifically evaluated on selected material and joining techniques. Specifically, the plastic pipe welding sector has a demand for an ADR technology. Therefore the developed methods will foremost be evaluated on this type of joining. System validation will entail a library of typical flaws and will examine the integration of the proposed framework into existing inspection processes and devices. As part of this research work, the proposed system will undergo laboratory and field trials, which will provide additional feedback and opportunities for identifying research gaps and progressing this work further.
Applicants must have a minimum undergraduate Honours degree (UK 2:1 or better) or MSc (Merit or Distinction) in a relevant Science or Engineering subject from a reputable institution. Overseas applicants should also submit IELTS results (with an overall score 6.5 or higher) if applicable. More details on entry requirements can be found at: https://www.sheffield.ac.uk/acse/research-degrees/applyphd
This project is funded by NSIRC, TWI and The University of Sheffield. The studentship will provide successful Home/EU students with a stipend of £20k/year, for 3 years, and will cover the cost of tuition fees. Overseas applicants are welcome to apply, with total funding capped at £24k/year, for 3 years.
How to Apply
To apply please submit a PhD application using our online application system via the Apply link at the following: http://www.sheffield.ac.uk/postgraduate/research/apply/applying
In your application, please state the project title and also indicate Dr George Panoutsos as your preferred supervisor.
Please note that deferral to a later start date for these Studentships is not permitted.
Deadline for Applications: 24 January 2018
Informal Contact: Dr George Panoutsos (firstname.lastname@example.org)