Project Code: 34727
Start date and planned duration: January 2022, 24 months
Objectives
This project aims at investigating the application of machine learning methods and conventional signal processing techniques to adaptively address the adverse impacts of high temperature during welding process. To achieve this goal, a test rig and prototype system will be developed to facilitate experiment and ultrasonic inspection under 450°C. Ultrasonic A-scans and transient temperature profiles are to be recorded for post analysis and algorithm development.
Project Outline
The project consists of three work packages, namely:
Work Package 1 – Literature research on state-of-the-art. The aim for this work package is to develop knowledge regarding the state-of-the-art of correction methods for high temperature UT applications.
Work Package 2 – High temperature experiment and data collection. In this work package, current development of the high temperature test rig will be improved in terms of couplant delivery, leakage prevention and data collection system optimisation, to enable high temperature UT capabilities for arc welding, the targeted temperature will be 450°C.
Work Package 3 – Comparative study of correction methods for high temperature UT application. In this work package, comparative study will be conducted using the UT data collected from the experiment. Both machine learning techniques and conventional image processing techniques will be investigated to develop an adaptive algorithm to address the high temperature impacts on the UT data.
Industry Sectors
- Power
- Oil & Gas
- Manufacturing
Benefits to Industry
Upon successful delivery of the project, the expected benefits include:
- Reliable ultrasonic inspection techniques for Arc Welding at elevated temperature conditions
- Early detection of indications will reduce cost, improve quality and assist scheduling
- A prototype system capable of ultrasonic inspection at high temperature which is able to be adapted for validation and further development work
- Investigation results on using Machine Learning technologies for Non-Destructive Evaluation.