Electron beam welding is used for fabricating critical components for the aerospace and nuclear industries which demand high quality. The cost of materials and associated processes of fabrication is also very high. Therefore, manufacturing processes in these industries are highly controlled. However, it has been found that even minor changes in the electron beam gun itself can produce large variations in beam characteristics, leading to unpredictable welding performance. Hence, it is very important to ensure the beam quality prior to carrying out welds. This requires some kind of device and process to characterise the electron beam to indicate variations. A detailed review of different technologies used to develop devices to characterise electron beams has been carried out. At this time, it is uncommon for beam measurement to be carried out on production EBW equipment. Research carried out for this thesis is focused on development of a novel approach to characterise the electron beams using a slit-probe to maintain the quality of the welds. The challenge lies in deriving relevant features from the acquired probe signal which can effectively differentiate between the beams of different quality. Wavelet transformation, with its advantages over other methods for simultaneous time and frequency localization of signals, has found its application to feature extraction in many pattern based classifications. This technique has been used to analyse probe signals considering that different quality beams will possess unique signal profiles in the form of their distribution of energies with respect to frequency and time. To achieve the aim of the thesis, an experimental approach was used by carrying out melt runs on Ti-6Al-4V plates focusing on aerospace requirements, and varying beam properties and acquiring probe signals for all beam settings. Extracted features from the probe signals have been used in classification of the electron beams to ensure these will produce welds within the tolerance limits specified by aerospace standards for quality assurance. The features vector was compiled following statistical analysis to find the significant beam characteristics. By analysing the performance of classifier for different combination of parameters of the features vector, the optimum classification rate of 89.8% was achieved by using the parameters derived from wavelet coefficients for different decomposition levels. This work showed that the use of wavelet analysis and classification using features vectors enabled identification of beams that would produce welds out-of-tolerance. Keywords: Electron beam welding, probe devices, electron beam characterisation, quality assurance, wavelet transform, features vector, linear discriminant classifier, weld profiles, weld defects.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:699273 |
Date | January 2016 |
Creators | Kaur, Aman P. |
Contributors | Balachandran, W. |
Publisher | Brunel University |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | http://bura.brunel.ac.uk/handle/2438/13436 |
Page generated in 0.0019 seconds